Close
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Using self-reported memory ability to measure memory ability in older adults: A meaningful measure or an inapropriate shortcut?
(USC Thesis Other)
Using self-reported memory ability to measure memory ability in older adults: A meaningful measure or an inapropriate shortcut?
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type o f computer printer.
The quality o f this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely afreet reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back o f the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6” x 9” black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to
order.
UMI
A Bell & Howell Information Company
300 North Zeeb Road, Ann Arbor MI 48106-1346 USA
313/761-4700 800/521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
USING SELF-REPORTED MEMORY ABILITY
TO MEASURE MEMORY ABILITY
IN OLDER ADULTS:
A MEANINGFUL MEASURE OR AN INAPPROPRIATE SHORTCUT?
by
Kerry Parker Bumight
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment o f the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Gerontology and Public Policy)
December 1996
Copyright 1996 Kerry Parker Bumight
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 9720197
Copyright 1996 by
Bumight, Kerry Parker
All rights reserved.
UMI Microform 9720197
Copyright 1997, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA 90007
This dissertation, written by
....................................
under the direction of hsz. Dissertation
Committee, and approved by all its members,
has been presented to and accepted by The
Graduate School in partial fulfillment of re
quirements for the degree of
DOCTOR OF PHILOSOPHY
Dean of ite Studies
Ju n e 5, 1996
Date
DISSERTATION COMMITTEE
Chairperson
• V
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ACKNOWLEDGMENTS
This dissertation reflects the guidance o f a number of important people who I would like
to thank:
Todd Bumight, my best friend and constant source o f laughter. There is no one I
admire more than Todd and I am grateful for every single moment with him.
John and Betty Parker, two incredible people who have showered me with love
and encouragement for 26 years.
Dr. Liz Zelinski, my doctoral committee chair, mentor, and cherished friend. It
has been an honor to work under Liz and I hope I may someday instruct my
students in the way Liz has instructed me.
Dr. Eileen Crimmins, a role model from the very start whose insight and advice
were invaluable to the research.
Dr. Kate Wilber, living proof that it is possible to be an accomplished, hard-
minded academic and kind, warm-hearted person.
Dr. Tim Biblarz, my outstanding research methods professor, outside committee
member, and generous giver of encouragement.
Dr. Sandy Reynolds, cohort mate, and wonderful friend. Always a few miles
ahead, she constantly turned back to urge me on.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
ACKNOW LEDGM ENTS.........................................................................................
LIST OF TABLES ....................................................................................................
ABSTRACT..................................................................................................................
CHAPTER I: INTRODUCTION............................................................................
A. Background .................................................................................
B. Research Questions ....................................................................
C. Contribution to the Literature....................................................
D. Organization of the Dissertation ..............................................
CHAPTER II: THEORETICAL BACKGROUND & LITERATURE REVIEW
A. Theoretical Background...............................................................
B. The Role of Sampling .................................................................
C. The Inclusion or Exclusion o f Proxy Respondents ..................
D. Measurement of Self-Reported Memory A bility......................
E. Measurement of Objective Memory Perform ance....................
F. H ypotheses.....................................................................................
CHAPTER III: RESEARCH DESIGN......................................................................
A. Samples: Combining Breadth and D epth...................................
B. M easures.........................................................................................
C. Analysis Procedures......................................................................
. ii
. . v
. x
. 1
. 1
n
. J
.4
. 5
. 6
. 6
14
2'1
24
29
31
32
32
36
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER IV: THE RELATIONSHIP BETWEEN SELF-REPORTED AND
OBJECTIVE MEMORY ABILITY ...................................................................................... 53
A. D escriptives................................................................................................53
B. The Role o f Sampling ...............................................................................70
C. The Inclusion of Proxy Respondents .................................................... 83
D. Measurement of Self-reported Memory Ability ...................................88
E. Measurement of Objective Memory Ability ......................................... 93
CHAPTER V: ACCURATE VERSES INACCURATE SELF-APPRAISAL OF
MEMORY A B ILITY .............................................................................................................. 99
A. D escriptives................................................................................................99
B. Accurate Memory Appraisal, Under-Estimation, Over-Estimation 103
C. Toward an Under-Estimator and Over-Estimator P ro file ....................108
CHAPTER VI: DISCUSSION AND IM PLICATIO N S.....................................................113
REFERENCES ........................................................................................................................ 125
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF TABLES
Table 1. Sampling Strategies o f Empirical Studies on the Relationship Between
Self-Reported and Objective Memory A bility..............................................15
Table 2. Measures o f Self-Reported Memory Ability in Empirical Studies on the
Relationship Between Self-Reported and Objective Memory Ability . . . 26
Table 3. Measures o f Memory Performance in Empirical Studies on the
Relationship Between Self-Reported and Objective Memory Ability . . . 29
Table 4. A Comparison of the Measures Used in the LBLS and AHEAD .............. 42
Table 5. Sex Distribution of Respondents in the AHEAD and LBLS .....................54
Table 6. Age Distribution of the Respondents in the AHEAD and LBLS .............. 55
Table 7. Racial and Ethnic Compositon o f the AHEAD and LBLS Samples .... 55
Table 8. Years o f Education by Sex in the AHEAD and LBLS Data Samples . . . 56
Table 9. United States Region o f Residence .................................................................57
Table 10. Self-Reported Health by Sex in the AHEAD Data S e t ...............................58
Table 11. Self-Reported Heakh by Sex in the LBLS Data Set ...................................58
Table 12. Self-Reported Health by Age Groups in the AHEAD Data Set ................59
Table 13. Self-Reported Health by Age Groups in the LBLS Data S e t...................... 60
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 14. Self-Reported Eyesight by Sex in the AHEAD Data S e t...........................61
Table 15. Self-Reported Eyesight by Sex in the LBLS Data S e t............................... 62
Table 16. Self-Reported Hearing by Sex in the AHEAD Data S e t ...........................63
Table 17. Self-Reported Hearing by Sex in the LBLS Data Set ................................63
Table 18. Depression Groupings by Sex in the AHEAD Data Set ............................63
Table 19. Depressive Symptoms by Sex in the LBLS Data S e t............................. 65
Table 20. Means and Standard Deviations of the Comparable Measures in the
AHEAD and LBLS Data S e ts .......................................................................66
Table 21. Intercorrelations Between Variables in the AHEAD Data S e t ..................68
Table 22. Intercorrelations Between Variables in the LBLS Data Set .....................69
Table 23. Self-Reported Memory Ability by Sex in the AHEAD Data S e t .............70
Table 24. Self-Reported Memory Ability by Sex in the LBLS Data S e t ................ 71
Table 25. Self-Reported Memory Ability by Age Group in the AHEAD Data Set 72
Table 26. Self-Reported Memory Ability by Age Group in the LBLS Data Set... 72
Table 27. Self-Reported Memory 1 Year Comparison by Sex in the AHEAD Data
Set .....................................................................*.............................................73
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
v ii
Table 28
Table 29.
Table 30.
Table 31.
Table 32.
Table 33.
Table 34.
Table 35.
Table 36.
Table 37.
Table 38.
Table 39.
Self-Reported Memory 1 Year Comparison by Sex in the LBLS Data
Set ..................................................................................................................... 73
Mean Proportion Correct in List Recall in the AHEAD and LBLS Data
S e ts......................................................................................................................74
Mean Proportion Correct in List Recall by Age in the AHEAD and LBLS
Data Sets ...........................................................................................................75
Results of the Model I Hierarchical Regression Analysis o f the Predictors
of Objective Memory Ability in the AHEAD and LBLS Data Sets .... 78
Results of the Model 2 Hierarchical Regression Analysis o f the Predictors
of Objective Memory Ability in the AHEAD and LBLS Data Sets .... 80
Descriptives of the Full AHEAD Sample (n=6436) and the AHEAD Sub-
Sample (n=230) .............................................................................................. 8 1
Results of the Model 1 Hierarchical Regression Analysis o f the Predictors
of Objective Memory Ability in the Full AHEAD Sample, the LBLS
Sample, and the AHEAD Sub-Sample ........................................................ 82
Use of Proxy Respondents by Age Groups in the AHEAD Sample .... 83
Means and Standard Deviations o f the Characteristics o f Proxy
Respondents and Self-Respondents in the AHEAD S am p le..................... 84
Relation of Proxy Respondent to the Selected Participant in the AHEAD
S am p le............................................................................................................... 86
Self and Proxy Ratings of Memory Ability for Each Response Option . 87
Proxy Rating o f Memory Ability by Proxy Relationship ..........................88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 40. Results o f the Model 1 Hierarchical Regression Analyses o f the Predictors
o f Memory Performance Using Abbreviated and Full Measures of Self-
Reported Memory A b ility ...............................................................................91
Table 41. Results of the Model 2 Hierarchical Regression Analyses of the Predictors
o f Memory Performance Using Abbreviated and Full Measures o f Self-
Reported Memory A b ility .............................................................................. 92
Table 42. A Comparison of the Measures o f Memory Performance in the LBLS
Data Set ............................................................................................................ 94
Table 43.
Table 44.
Table 45.
Table 46.
Table 47.
Table 48.
Table 49.
Results of the Hierarchical Regression Analyses of the Predictors of the
Abbreviated and More Extensive Measure of Objective Memory
Ability ...............................................................................................................95
Results of the Hierarchical Regression Analyses of the Predictors of the
Abbreviated and More Extensive Measure of Objective Memory Ability
Using Abbreviated and Extensive Measures of Self-Reported Memory'
Ability ...............................................................................................................97
Results o f the Model 2 Hierarchical Regression Analyses of the Predictors
o f the Abbreviated and More Extensive Measure of Objective Memory
Ability Using Abbreviated and Extensive Measures of Self-Reported
Memory A b ility ................................................................................................98
A Model of the Accuracy of Self-Ratings of Memory’ Ability .................99
Recoding the AHEAD variables into 3 Category V ariab les....................101
Self-Rated Memory Ability by Objective Memory Performance in the
AHEAD Data Set..............................................................................................102
Self-Rated Memory Ability by Objective Memory Performance in the
LBLS D ataS et.................................................................................................103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 50. Accurate Assessment, Under-Estimation, and Over-Estimation in the
AHEAD and LBLS Samples...........................................................................103
Table 51.
Table 52.
Table 53.
Table 54.
Table 55.
Table 56.
Results of the Multinomial Logistic Regression Analysis Predicting the
Probability of Under-Estimating and Over-Estimating Memory Ability
Compared to Accurately Rating Memory Ability....................................... 107
Characteristic Profiles Associated with an Increased Probability o f Under-
Estimating and Over-Estimating Memory Ability.................................... 108
The Probability of Under-Estimating Memory Ability Under Seven
Hypothetical Vignettes Each Manipulated to Highlight a Single
Characteristic.................................................................................................. 109
The Probability of Over-Estimating Memory Ability Under Seven
Hypothetical Vignettes Each Manipulated to Highlight a Single
C haracteristic.................................................................................................. 110
A Profile of the Characteristics of An "Under-Estimator" and an "Over-
Estimator" Ranked by Order of Importance ..............................................I l l
A Comparison of the Characteristics Associated with Under- and Over-
Estimating Memory Ability and the Characteristics Associated with Better
or Worse Memory Performance.................................................................... 121
with permission of the copyright owner. Further reproduction prohibited without permission.
Using Self-reported Memory Ability to Measure Memory Ability In Older Adults:
A Meaningful Measure or an Inappropriate Shortcut?
Growing interest in the cognitive functioning o f America's senior population has
moved research on memory ability from the exclusive domain o f cognitive research into
the multidisciplinary arena of epidemiogical research. Interested in gauging the memory
ability o f respondents, but unable to conduct a full memory assessment because of the
time constraints associated with large sample, investigators have included measures of
self-rated memory ability in the cognitive functioning sections o f their survey
instruments. The aim o f this research was to determine: (1) whether self-reported
memory ability accurately reflects objective memory ability; and (2) whether there are
systematic differences between those who are accurate in their memory appraisals and
those who are inaccurate. To avoid the methodological weakness that have compromised
previous investigation in this area, the research design attempted to combine the breadth
of social epidemiology with the depth o f psychological inquiry. To this end. two data sets
were used: a nationally representative epidemiological survey o f 6436 adults ages 70-103
(AHEAD); and a convenience sample o f 230 adults ages 70-95 recruited for an extensive
cognitive testing (Long Beach Longitudinal Study). Using the strengths of the two data
sets, analyses were conducted to examine the predictive utility o f self-reported memory
ability and the characteristics associated with under-estimating and over-estimating
memory ability. Results indicated that self-reports o f memory ability are only modestly-
related to objective memory performance. There were systematic differences between
those who accurately appraised memory ability as opposed to those who were inaccurate.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Specifically, under-estimating memory ability was associated with being younger, more
educated, white, and reporting a more negative sensory status. In contrast, those who
over-estimated memory ability were more likely to be older, male, less educated, black,
and report a more positive sensory and health status. Taken together, the results indicate
that self-reported memory ability is an inappropriate shortcut for measuring memory
ability. The characteristics associated with better memory performance are associated
with under-estimation, and inversely related to over-estimation. Future research is
needed to explore the relationship between cognitive impairment and over-estimation and
performance expectation and under-estimation of memory ability.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER I: INTRODUCTION
A. BACKGROUND
Self-rated memory ability is becoming an increasingly common element of
surv eys o f the aged. Interested in gauging the memory ability o f respondents, but unable
to conduct a full memory assessment because o f the time constraints associated with large
samples, researchers have included measures of self-rated memory in the cognitive
functioning sections of their survey instruments. Influential epidemiological surveys
such as the National Health Interview Survey (NHIS), the Health and Retirement Survey
(HRS), and the study o f Asset and Health Dynamics Among the Oldest Old (AHEAD)
include measures of self-reported memory ability. Despite the popularity o f the measure,
interpretation o f self-rated memory ability is controversial because it is unclear whether
self-rated memory ability accurately reflects objective memory ability. Empirical
research on the relationship between self-reported memory ability and objective memory
performance has lead to discrepant conclusions with some supporting, and others
refuting, the predictive utility of self-reports o f memory ability. The discrepant findings
stem from methodological shortcomings inherent in the designs of the empirical research
on the predictive utility of self-reported memory ability.
The aim of this dissertation is to determine whether self-reported memory is a
meaningful measure, or inappropriate shortcut, for measuring memory' performance. To
achieve this aim. the research design must overcome the methodological shortcomings
that have prevented conclusive findings on the relationship between self-reported
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
memory ability and objective memory performance. These shortcomings can be grouped
into two areas: inadequate breadth and inadequate depth. Inadequate breadth refers to the
small and unrepresentative samples used in the empirical literature on the relationship
between self-reported and objective memory ability. O f twenty-two frequently cited
studies on the self-rated and objective memory ability, only two studies had samples over
200. and the mean sample size was 134 participants. Inadequate breadth compromises
the external validity of the findings and conclusions drawn from unique samples cannot
be generalized to the rest of the population. Perhaps even more disconcerting than the
size of the samples, is their composition. Most samples have been recruited in one of
three ways: offering memory assessment and training through newspaper advertisements:
using outpatients of depression or psychiatric clinics; or recruiting convenience samples
of subjects willing to participate in research on aging. All three recruitment strategies are
likely to attract participants that are systematically different from the general population.
A second factor compromising the external validity of the study stems from the
fact that in the senior population, there are older adults who are unwilling (too busy or
disinterested) or unable (too physically and/or cognitively impaired) to answer the types
of questions used in cognitive research. In order for a sample to represent the population,
a sample must include those who are unwilling, or unable, to complete the test protocol.
In other areas o f gerontological research this has been achieved through the use o f proxy
respondents who provide information on the selected participants. Unfortunately,
research on the predictive utility of self-reported memory has not used proxy respondents
as part o f its sampling strategy.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Turning next to the problems associated with inadequate depth, the relationship
between subjective and objective memory ability has been obscured by the measures used
to operationalize self-reported memory ability and objective memory performance. In the
majority o f studies on self-reported memory ability, the internal validity has been
compromised by the use o f measures of self-reported memory which have not been
established to be psychometrically reliable.
B. RESEARCH QUESTIONS
Much o f the literature on the relationship between self-reported and objective
memory is based upon research studies with small, biased samples (inadequate breadth):
and/or employ poor measures of the central constructs self-reported memory ability and
memory performance (inadequate depth). As a result o f these shortcomings in the
empirical literature, two fundamental questions remain: (1) Do self-reports of memory
ability reflect objective memory ability? (2) Are there systematic differences between
those who accurately appraise their memory ability as opposed to those who over- or
under-estimate their ability? In approaching these questions, attention will be paid to the
role of sampling, research design, and measurement in an attempt to avoid the
methodological pitfalls that have compromised previous research.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4
C. CONTRIBUTION TO THE LITERATURE
The strength o f gerontology has derived from the willingness o f scholars to cross
the boundaries of their own training in order to pursue a broader and deepening
understanding of aging. The growth o f an intellectually rich gerontology depends on the
willingness to foster interaction between disciplines (Cole, 1995). With the 1994 release
o f the Asset and Health Dynamics o f the Oldest Old (AHEAD) data set. there is an
unprecedented opportunity to combine the breadth o f social epidemiology with the depth
of psychological inquiry to examine the predictive utility of self-reported memory ability.
Never before have researchers been able to examine the relationship between self-
reported memory ability and objective memory' performance using a sample that
approximates the U.S. population. Until now, findings on the predictive utility o f self-
reported memory ability have been derived from cognitive studies drawn from small, and
generally unrepresentative, convenience samples. However, the breadth offered by the
scale o f the AHEAD study is not without cost. The measures used to operationalize self-
reported memory' ability and objective memory performance are considerably abbreviated
as compared to their counterparts in cognitive research. In order to determine how the
internal and external validity o f the research design impacts the appearance of the
relationship between self-reported and objective memory ability, it is now possible to
incorporate the breadth o f population research with the depth o f cognitive research. The
findings of this research will inform scholars on the appropriate interpretation of self-
reported memory ability and on the advisability of its continued inclusion in
epidemiological research. If self-reported memory ability is a meaningful measure of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
memory ability, its inclusion in the cognitive functioning section o f population surveys
may be warranted as a feasible and economically sound way to approximate memory
ability. If. however, self-reported memory ability does not reflect objective memory
ability, scholars may need to reconsider its inclusion in survey research and instead focus
upon a more costly, but accurate, measure o f memory ability. Accurate assessment and
interpretation o f research constructs are of utmost importance not only for the research
and policy community, but ultimately to the community at large who will benefit from
accurate information and policy decisions based upon that information.
D. ORGANIZATION OF THE DISSERTATION
The dissertation contains five chapters, the first consists o f a general introduction
and a discussion of the impetus behind the research. The second chapter reviews the
theoretical and empirical literature on the relationship between self-reported and objective
memory in light of the methodological factors that have influenced the appearance of the
relationship between the two constructs. This review is then used to generate two
hypotheses that will be tested in the research. Chapter Three describes the methodology
by discussing the two samples employed, their respective measures, and the analysis
procedures used to test the research hypotheses. Chapter Four presents the results from
the analyses used to test hypothesis one. and Chapter Five reviews the results o f the
hypothesis two tests. The final chapter. Chapter Six, discusses the conclusions,
implications, and limitations of the research and proposes several recommendations based
on these findings.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6
CHAPTER II: THEORETICAL BACKGROUND
AND LITERATURE REVIEW
In order to avoid collapsing under the weight of disparate empirical findings on
the relationship between self-reported memory ability and objective memory ability, a
systematic approach is needed to provide an organizational framework within which
observations can be assimilated. To this end, this chapter begins by clarifying the area of
interest by framing self-reported memory ability in the theoretical work surrounding the
topic. The next section examines the literature on the correlates o f self-reported and
objective memory ability. Third, the relationship between self-reported memory ability
and objective memory performance is discussed in light of the methodological issues that
have shaped the outcomes. The final section articulates the hypotheses that are generated
from the literature review.
A. THEORETICAL BACKGROUND
The extent to which people are aware o f their own memory ability has long been a
subject o f interest to psychologists and philosophers. Cavanaugh and Perlmutter (1982)
point to eighty years o f scholarly inquiry on memory awareness and self-appraisal. When
considering the awareness of memory ability, the term awareness refers to a state marked
by realization, perception, or knowledge (Klatzky. 1984). Awareness may indicate either
general information, wide knowledge, interpretive power, or vigilant perception.
Roberta Klatzky (1984) offers a useful conceptualization of memory awareness that is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7
grounded in the information processing approach to human cognition. The basic tenet of
information processing theory is that the human memory system can be seen as a set of
structures representing information and a set of processes acting on that information.
Klatzky postulates that memory awareness is composed o f three distinct, yet interrelated,
categories: on-line memory awareness, epistemic memory awareness, and personal
memory models.
On-line memory awareness is the awareness o f ongoing memory processes. It
concerns the awareness o f currently occurring perceptual, cognitive, and motoric
activities. In everyday use, this type of awareness is often expressed as being conscious
of something. Because the term consciousness carries with it many connotations, both
mystical and psychoanalytical, Klatzky's on-line awareness is more fitting when referring
to information processing . In the literature, on-line awareness of memory functioning
has been examined under the term memory monitoring, or more generally, as executive
processing (Cavanaugh. 1989). On-line awareness has generally been examined in two
ways: (1) by asking people to predict how well they think they will perform on a task
prior to actually doing it: or (2) by asking people to predict how well they think they did
after attempting the task (e.g. studying the list). With some exceptions (Berry. West. &
Scogin. 1983: Camp. Markley. & Kramer, 1983), studies examining the accuracy of
predictions made prior to performance show that older adults tend to overestimate how
well they will do (Bruce. Coyne. & Botwinick, 1982: Murphy. Sander. Gabriesheski. &
Schmitt. 1981). Studies that ask subjects to make assessments after attempting the task
demonstrate that older adults are as accurate in predicting their recall and recognition
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
8
performance as younger adults (Bruce, Coyne, & Botwinick, 1982). However, in both
groups, these judgements o f learning are significantly more accurate when assessed
shortly after, rather than immediately after, the task has been attempted. (Dunloskv &
Nelson. 1995). The difference in the accuracy o f judgements of learning is thought to
reflect transient information that affects immediate judgements. Judgements o f learning
that are slightly delayed until the transient information has dissipated, have been shown
to predict performance more accurately (Thiede & Dunlosky, 1994).
Based on the word epistemology. the theory of knowledge. Klatzky calls the
second type of memory awareness epistemic awareness. This type of memory awareness
refers to the awareness of past experiences that are stored in memory. Sometimes we are
aware of the contents of memory because we can revive, or to use information processing
wording, "retrieve" them. Other times, retrieval may be impossible, but we can still have
a feeling about what we know based upon other relevant information that can be
retrieved. For example, you may be aware that the name of a high school classmate is in
your memory even though you are not able to say what the name is right now. Epistemic
awareness then, is the awareness of the inventory of what our memory holds. Epistemic
awareness includes awareness o f what you know, how you came to know it. and how
well you know it.
Much of the research on epistemic awareness has focused upon the experience of
being unable to recall something yet being aware that we know it. William James (1890)
describes this state as being "beckoned in a given direction, making us at moments tingle
with the sense of our closeness and then letting us sink back without the longed for term
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(p.251)." Hart (1965) studied the "feeling o f knowing" by presenting participants with a
task such as "name the largest planet" and then asking them to indicate the feeling of
knowing for any items that could not be recalled. When subjects were presented with the
answers in the recognition test. Hart observed that subjects were accurate in their feelings
of knowing. When subjects indicated that they felt they knew unrecalled items, they were
more likely to recognize the correct answer relative to the unrecalled items that they felt
they did not know. Research on epistemic awareness also documents that we can retrieve
things that are related to what we cannot currently remember. Brown and McNeill (1966)
encouraged Tip o f the Tongue (TOT) states in college students by giving them words that
were sufficiently obscure as to be hard to recall, but common enough that they were
likely to be familiar with them. Students could report things about the words (number of
syllables, initial letter, other words with familiar sound or meaning), even when they
could not report the word. Similarly, when Bruce et al. (1982) asked older adults to
recall the names o f obscure entertainers, the older adults could recall related information
even when the words were halted "on the tip o f their tongue". Other applications of
epistemic awareness that have received attention in cognitive developmental research
include eyewitness accounts, and the difference between perceived and generated
memories (Cohen & Faulkner. 1989).
The third type of memory awareness, personal memory models, refer to the
articulable beliefs that an individual holds about the general nature of human memory' and
his or her particular capabilities (Klatzky. 1984). This concept has also been termed
Metamemory (Flavell. 1971) and is most commonly defined as one's knowledge.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10
perceptions, and beliefs about the functioning, development, and capacities o f one's own
memory and the human memory system (Dixon, 1989). There is confusion surrounding
the terms metcimemory and personal memory models because the breadth o f the
definition has spawned multiple, and sometimes mutually exclusive, operational
definitions. It appears that the confusion stems from the existence o f two separate
concepts subsumed under a single term. The situation could be improved by separating
the two constructs grouped under the term personal memory models or metamemory.
The two concepts are: (1) the awareness or knowledge that an individual holds about how
human memory works in general; and (2) the awareness or beliefs that an individual
holds about his particular memory ability. The first concept could called systemic
memoiy awareness, and the second called metamemory. These terms have been used to
capture the concepts before, but have never been explicitly specified in this way. Using
this more refined conceptualization. Klatzky's typology would be expanded into four
types o f memory awareness: on-line, epistemic. systemic. and metamemory. These
concepts can be summed up by their corresponding questions: How am I doing with
remembering this? (on-line): What do I have stored in my memory? (epistemic): What is
human memory capacity like? (systemic); and How sound is my own memory ability
(metamemorv)?
Here, systemic awareness is the awareness or knowledge that a person holds about
how memory works in general. In his work on on-line awareness. Cavanaugh (1989)
used the term systemic awareness to refer to general knowledge about memory but then
also included the second, and I argue separate, concept of awareness about one's own
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
11
memory ability. As defined here, systemic memory awareness is what a person knows
about human memory. For example, most people know that recall is typically harder that
recognition, that short-term memory is limited, and that mnemonic devices can help
people remember things. Perlmutter (1978) studied this concept under the term
metamemorial knowledge and found no age differences between younger and older
adults.
The fourth type o f memory awareness, metamemory, is the concept most central
to this research. Under the more precise definition used here, metamemory refers to the
awareness or beliefs that an individual holds about his own memory ability. It is from
metamemory awareness of memory ability that people are able to articulate about, and
appraise, their memory ability. Self-reports o f memory have received considerable
attention from psychologists interested in awareness o f cognitive functioning (Dixon &
Hultsch. 1983b) and the accuracy of memory complaints (Zelinski. Gilewski. &
Thompson. 1980: Bolla. 1991). Scholars in this area theorize that the awareness or
beliefs that an individual holds about her own memory ability is composed o f various
dimensions. While numerous dimensions have been suggested, and subsequently used to
operationalize self-reported memory ability, there seems to be convergence on several
key dimensions including: frequency of forgetting, perceived change across time,
seriousness o f memory failures, strategies and mnemonics usage, memory knowledge,
demands on memory in daily life, memory for past events, and effort when forgetting
occurs (Gilewski & Zelinski. 1985).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Within the last decade, metamemory has attracted a wider audience including
sociologists and demographers interested in the feasibility of using self-reports of
memory ability as some type o f representation o f memory ability. Using the 1984
Supplement on Aging to the National Health Interview Survey (n=14.783). Cutler and
Grams (1988) found that 56% o f the adult population age 55 and over reported that they
have "trouble remembering things”. Specifically. 17% say they frequently had trouble.
39% report that they sometimes had trouble. 19% say they rarely had trouble, and 20%
indicate they never have trouble. Unfortunately, die National Health Interview Survey
(NHIS) did not include any measure of objective memory performance and thus the
findings, though important as the first o f their kind, only provide information on the
correlates o f memory complaints but not on their accuracy. Studies on the correlates o f
self-reported memory ability, have generally found that age, sex. education, health status,
and depression correlate with self-reported memory ability. Being older is associated
with more negative self-ratings o f memory ability. In the NHIS sample. 80% of
respondents aged 85 and over reported some problems, compared with 59% o f the
youngest group in the sample, ages 55-59 (Cutler & Grams. 1988). While increasing age
is associated with lower scores on objective memory performance, there is also evidence
that old people expect to have memory problems (Erber. Szuchman. & Rothberg. 1990)
and thus are more aware of any memory shortcomings (Berry. 1986). Women give
themselves more negative self-ratings of memory ability than men even though they out
perform men on memory performance tasks (Bolla et al., 1991). Verbrugge's (1987)
hypothesis that women may be more willing to report symptoms of illness, may apply.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Having fewer years o f education predicts a lower self-rating o f memory status. In
the NHIS study, the percentages o f people who reported having frequent problems
remembering was significantly higher for those who had less education, as was
percentage o f people reporting memory decline in the last year. Health variables have
been shown to predict self-reported memory ability. In fact, in the NHIS. the best
predictors o f self-reported memory are self-assessed health, functional, and sensory
problems (Cutler & Grams. 1988). Adults who report a worse health status, more
functional limitations, more vision impairment, and more trouble with hearing, tend to
report a more negative self-rating o f memory ability (Bazaragan & Barbre. 1994).
Virtually every study examining depressive affect and complaints has reported a
relationship indicating that a greater degree of complaint is associated with a greater
degree o f depression (Zelinski & Gilewski, 1988). The relationship between depression
and objective memory is less clear, with some studies concluding that older adults who
are more depressed do worse on memory tests, and others concluding that depression is
not associated with worse memory performance. Using meta-analvtic techniques to
synthesize data from 99 studies on recall and 48 studies on recognition in clinically
depressed and non-depressed samples. Burt. Zembar, and Neiderhe (1995) found a
significant, stable association between depression and memory impairment. However,
the authors qualified the results suggesting that depression is linked to particular aspects
of memory and that the linkage is found in particular subsets of depressed individuals.
Having considered the correlates of self-reported memory ability, the next section
considers what is known about the relationship between self-rated memory ability and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
objective memory performance. Here, the literature is somewhat more contradictory and
tangled with methodological shortcomings. It is therefore useful to examine the literature
on the predictive utility of self-reported memory ability in light o f the methodological
issues that have impacted the results. The most central methodological issues include:
( I ) the role o f sampling; (2) the inclusion or exclusion o f proxy respondents: (3)
measurement of self-reported memory ability; and (4) the measurement of objective
memory performance. The following sections review the empirical literature on the
relationship between self-reported and objective memory ability in light of these
methodological factors in order to determine how they have shaped the appearance of the
relationship between subjective and objective memory.
B. THE ROLE OF SAMPLING
Evidence on the relationship between self-reported memory ability and objective
memory performance has been derived laboratory and clinical studies based on small,
unrepresentative samples. Samples have generally been recruited in one of three ways:
offering memory assessment and training; using outpatients o f depression/psychiatric
clinics: and recruiting subjects to participate in cognitive research studies. Table 1
summarizes the sampling strategies of 22 frequently cited empirical studies on the
relationship between self-reported memory ability and objective memory' ability. The
third column, labeled "Outcome ”, summarizes the findings o f each study by using a "+ ‘*
or to indicate whether the results of the study support the predictive utility of self-
reported memory ability.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
15
Table I. Sampling Strategies of Empirical Studies on the Relationship Between
Self-Reported Memory Ability and Objective Memory Performance.
Study Sample Outcome
Zelinski et al. (1980) 196 Student/senior volunteers, Ages 18-80 +
Reige (1982) 60 Hospital staff volunteers. Ages 21-84 +
Dixon & Hultsch (1983a) 318 Female volunteers, Ages 18-84 +
Zelinski et al. (1990) #1 198 Volunteers for IQ study. Ages 55-85 +
Zelinski et al. (1990) #2 89 Volunteers for assessment Ages 50-87 +
Christensen (1991) 64 Volunteers for memory clinic. Ages 17-79
4-
Bassett et al. (1993) 810 Patients of psychiatric clinic. Ages 18-92 +
Fingerman et al. (1994) 151 Volunteers for clinic. Ages 20-85 +
Williams et al. (1987) 50 Referred for depression Ages 40+
4 -
Zarit (1982) 79 Volunteers for memory clinic, Ages 50-82
4-
Kahn et al. (1975) 153 Outpatients at a psych, clinic. Ages 50+ -
W estetal. (1984) 67 Women in memory clinic. Ages 65-90 -
Scogin (85) 59 Volunteers for memory clinic. Ages 60-82 -
Sunderland et al. (1986) 60 Particpants in Aging Research. Ages 64-75 -
O'Hara et al. (1986) 77 Patients from depression referral. Age 65+ -
Popkin et al. (1982) 41 Participants, Ages 50+ -
Bolla & Bleecker (1986) 17 Dr. Referrals for memory eval. Ages 40+ -
Poitrenaud (1989) 125 French Mangers. Ages 63-64 -
Zonderman et al. (1989) 58 Complainers from BLSA, Ages 50+ -
Scogin & Rohling (1985) 55 Volunteers for memory clinic. Ages 60-88 -
Bolla et al. (1991) 199 Volunteers, Ages 39-89 -
Hanninen et al. (1994) 20 Volun w/ memory complaints. Ages 67-78 -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
16
Recruiting subjects by offering memory training has been used in a number of
studies on the predictive utility of self-reported memory ability (Zarit. 1982; Scogin.
1985: West. Boatright. & Schlesser, 1984; Bolla-Wilson & Bleecker. 1986; Christensen.
1991). Scogin (1985) asserts that this technique results in "a more clinically relevant
population than randomly selected older adults"( p.81) because these are the individuals
who are more concerned about their memory abilities. However, this method is only
satisfactory if the researchers are only interested in the predictive utility o f self-reported
memory in the relatively few individuals who respond to advertisements offering memory
training. The results o f studies using this recruiting strategy cannot be generalized to a
broader population because those who respond to newspaper advertisements offering
memory training are likely to be systematically different than those who do not respond.
Comparing a group o f older adults who were recruited by offering memory training to a
group recruited to participate in research on aging. Zelinski et al. (1990) found that
although the groups did not differ in levels o f self-reported depression or memory
performance, the group who responded for memory training had significantly more
negative self-assessments of memory ability. Memory assessment may attract
individuals who are especially concerned about their memory abilities. For some
individuals these concerns may be well-founded, but the literature suggests that for
others, these concerns may reflect inaccurate and exaggerated concerns that reflect
anxiety.'hypochondriacal tendencies, or the tendency to complain. Using the Minnesota
Multiphasic Personality Inventory (MMPI) to measure hypochondriasis. Hanninen et al.
(1994) found that those who had higher scores on the Hypochondriasis scale had
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
17
significantly more negative self-rating of memory ability after age. sex. education, and
depression were controlled. Similarly, Zonderman et al. (1989) found that for those who
report memory complaints in the absence of memory problems, complaints reflect
tendencies toward somatic complaining. If people who respond to advertisements
offering memory training are more concerned, anxious, or hypochondriacal than those
who do not respond, then studies that use this recruiting method would be more likely to
find little or no association between self-rated memory ability and objective memory
performance. West. Boatwright, and Schleser (1984) recruited a group of 67 women
aged 65-90 by offering a memory training program. The self-reported memory measure
consisted of a 62-item questionnaire and the memory measures included list recall,
related list recall, digit span, and related digit span. Affective status was measured using
the Beck Depression Inventory and the State Anxiety Inventory. The author's concluded
that anxiety and depression are more closely related to self-reported memory ability than
objective memory performance. But if samples seeking memory training are more
anxious or depressed than the general population, these conclusions cannot be applied to
anyone beyond the sample group. This phenomenon is also apparent in a study by
Scogin (1985) who recruited 59 subjects by offering a memory training program, and
another 25 by recruiting individuals for research participation. The subjects recruited for
memory training had significantly more negative self-assessments on the Metamemory
Questionnaire than the subjects recruited for a research study. Those seeking memory
training were found to evidence lower levels o f correlation between self-reported memory
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
18
ability and memory performance on list recall and digit span tasks, than were the elders
not actively seeking memory training.
The second strategy, recruiting subjects from psychiatric/depression outpatient
clinics, has also been used in a number o f empirical studies (Bassett & Folstein. 1993:
Kahn et al..!975: O'Hara et al.. 1986; Popkin et al.. 1982). This strategy shares the same
weakness that offering memory training does, namely, unique samples. Those in
treatment at a psychiatric or depression outpatient center may be more depressed and/or
more neurotic than those who are not in treatment. This is significant given the strong
evidence on relationship between depression and self-reported memory ability and the
budding literature on the relationship between neuroticism and self-reported memory
ability. Research examining depressive affect and self-reported memory indicates that a
more negative self-rating o f memory ability is associated with a greater degree of
depression. This finding has been replicated for complaint scales developed for
individual studies (West. Boatwright, & Schleser. 1984; Larrabee & Levin. 1986): for
studies using the Short Inventory of Memory Experiences (SIME) (Erber. Szuchman. &
Rothberg. 1990): and for studies using the Memory Functioning Questionnaire (MFQ)
(O'Hara. Hinrichs. Kohout. Wallace & Lemke. 1986: Williams. Little. 8 c Scates. 1987:
Bolla. Lindgren. Bonaccorsy. & Bleecker. 1991). The measures of depression used in
these studies include: the Center for Epidemiologic Studies Depression Scale, the Beck
Depression Inventory, and the Geriatric Depression Scale. Regardless of inventory, most
studies conclude that memory complaints in community dwelling older adults are related
to depressed mood. Indeed, a number of studies suggest that self-reported memory is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19
more strongly associated with depression than with performance on memory tests (Bolla
et al.. 1991).
Guided by the association between affective states and self-reported memory,
scholars have also identified a relationship between the trait, neuroticism. and self-
reported memory ability (Poitrenaud et al.. 1989; Zonderman. Costa. & Kavvas. 1989:
Bassett & Folstein. 1993: Hanninen et al.. 1994). This focus broadens the inquiry by-
hypothesizing that self-reports of memory ability are not only impacted by one's present
state of depression, but also by the enduring trait characteristic, neuroticism.
Neuroticism refers to an individual’ s level o f emotional adjustment and instability. This
trait is characterized by the general tendency to experience fear, embarrassment, anger,
anxiety, sadness, guilt, self-consciousness, and disgust. As an outgrowth o f these
disruptive emotions, individuals high in neuroticism are also prone to impulsivitv.
vulnerability, and irrational ideas (Costa & McCrae, 1989).
Bassett and Folstein (1993). studied 810 subjects age 18-92. nearly half o f whom
were recruited from a psychiatric outpatient clinic. Grouped by major diagnostic
divisions: 139 subjects had anxiety disorders. 47 had affective disorders. 13 had
schizophrenia. 62 had cognitive disorders (dementias, delirium, mental retardation, and
developmental delay). 55 had substance abuse disorders. 21 has adjustment disorders,
and 36 multiple disorders. Self-reported memory was measured using the yes/no
question. "Do you fin d that you have trouble with your memory’?" With the exception of
subjects with substance abuse disorders, subjects with disorders were more likely to
respond "yes' than were those with no diagnosis. Subjects with substance abuse disorders
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
20
complained less o f memory problems than those with no diagnosis. Subjects with
schizophrenic, anxiety, and adjustment disorders exhibited greater discrepancies between
self-reported memory ability and memory performance. The investigators suggest that
the findings indicate the importance of emotional distress as a predictor of self-reported
memory, but not o f actual memory performance, for a number o f psychiatric categories.
The problem is that studies that use outpatient samples cannot conclude that neuroticism
and somatic complaining generally underlie memory complaints in the general
population. If being more neurotic is associated with more negative self-reports of
memory ability, a sample from an outpatient clinic that is composed of people high in
neuroticism would be unlikely to exhibit a relationship between self-reported memory
ability and objective memory performance. In a sample that is not overwhelmingly
composed of psychiatric outpatients, negative self-reports o f memory ability may be
accurate predictors of lower performance on objective measures.
The third strategy, recruiting subjects to participate in cognitive research on aging
has been the most common sampling technique in studies on the predictive utility o f self-
reported memory ability (Zelinski etal, 1980; Dixon & Hultsch. 1983; Zelinski et al..
1990; Sunderland et al.. 1986; Scogin & Rohling. 1989; Bolla etal.. 1991). While
perhaps less blatant, this strategy is also subject to response bias in that those who agree
to complete the extensive testing protocol involved with cognitive research, are likely to
be systematically different than those who refuse to participate. Participation in cognitive
research often requires travel to the testing site, adequate vision and hearing, and stamina
for several hours o f testing. For this reason those who are in poor health, have sensory
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
trouble, or are cognitively impaired are generally excluded from the samples. The
literature suggests that individuals with health problems, sensory problems, or functional
limitations in activities of daily living are more likely to feel that their memory had
deteriorated over the past year (Cutler & Grams. 1988). These health indicators may
contribute to memory complaint through their effect on affective state, or it may be that
those who are in worse health are. in fact, experiencing more trouble with memory
ability. It is also may be the case that medication use which can effect central nervous
system functioning, and in turn, cognition and mood, can exacerbate complaints (Zelinski
& Gilewski. 1988).
When prospective subjects are recruited in cognitive research, they are told that
the research entails tests of memory and cognitive functioning. Individuals who are
uncomfortable with their ability in this area because o f low education, or the perception
o f significant decline, may be more likely to refuse to participate. Having fewer years of
education predicts a lower self-rating of memory ability (Cutler & Grams. 1988). The
problem of inability or unwillingness to participate results in samples that are unique and
thus, ungeneralizable.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
22
Only recently have self-assessments of memory ability been included in large
scale epidemiological research. As previously discussed, the 1984 Supplement on Aging
o f the NHIS included two measures o f self-reported, everyday memory problems. But
because the survey did not include a measure of objective memory performance, it lends
no insight into the predictive utility o f self-reported memory ability. The AHEAD data
set is among the first population study to include measures of both subjective and
objective memory ability. However, because the measures o f self-reported memory
ability and objective memory performance are extremely abbreviated, there is some
question surrounding the reliability and validity of the measures. Later in this chapter,
this issue will be examined in depth in the discussion o f the role o f adequate
measurement of self-reported memory ability and objective memory ability.
C. THE INCLUSION OR EXCLUSION OF PROXY RESPONDENTS
Proxy respondents often play a crucial role in surveys targeting the elderly
because the very phenomena of interest to the survey researcher, such as diminished
physical or cognitive ability, makes it difficult to interview the older respondent directly.
A procedure followed by the AHEAD Survey, and most other large scale surveys o f the
aged, is to interview a person knowledgeable about the selected participant such as a
spouse or adult child. The need for proxy respondents in surveys of the elderly is
suggested by data from the National Long Term Care Survey which is a seven year, three
wave study of older adults. O f the 1.839 respondents who were eligible to participate in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
all three waves of data collection, approximately 44% had a proxy respondent provide
some, or all. of the information in one or more of the data waves. (Hurd et al.. 1994).
If proxy ratings o f memory ability differ systematically from self reports of
memory ability, then using proxy measures could introduce problems o f its own. The
literature on the adequacy of informant rating suggests that there may be age differences
in how people appraise memory problems. Using vignettes o f everyday memory
problems. Erber. Szuchman. and Rothberg (1990) found that the failures o f older people
were judged as signifying greater mental difficulty, and greater need of memory training
than were the identical failures o f young targets. Informants were found to rate memory
failures more stringently than self-respondents, and being a younger informant was
associated with harsher appraisals of memory problems. Among informants, it appears
that ratings given by spouse proxy respondents most closely resemble self-ratings.
Studies examining the adequacy o f spouse ratings of memory ability have found that the
informant ratings on the Memory Functioning Questionnaire (MFQ) approximate self-
ratings. and are as good, if not better at predicting objective memory ability than are self-
reports (Scogin & Rohling. 1989: Zelinski et al.. 1990). It is may be that he knowledge
that your spouse is also rating your memory ability, impacts the accuracy of your self-
rating (Zelinski et al.. 1990). It is important to note that the MFQ is considerably more
thorough than the single item used in survey research which may compromise the
accuracy of proxy reports (as well as self-reports).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
24
D. MEASUREMENT OF SELF-REPORTED MEMORY ABILITY
In part, the contradictory findings on the relationship between self-reported
memory and memory performance reflect the fact that most studies do not offer explicit
definitions o f either construct. Instead, definitions of the concepts are inextricably tied to
the operationalization o f the concepts. This is problematic given the lack o f consensus
surrounding the proper operationalization o f self-reported memory ability. Scogin
(1985) describes the study of older adults' perceptions and feelings about memory’
functioning as "fraught with terminological confusion" (p. 2). The terms self-reported
memory ability, memory complaint, memory evaluation, and metamemory have been
used almost interchangeably. Often the definition of self-reported memory is implicitly
derived from its operationalization which is problematic because most self-reported
memory indices have not been established to be psychometricaily reliable. Gilewski and
Zelinski's (1986) review of ten questionnaires frequently employed in research on self-
reported memory ability reveals that only two. the MIA and the MFQ. possess internal
structure and reliability considered acceptable by psychometric criteria.
The lack of clarity surrounding the definition and operationalization o f the
concept o £ self-reported memory ability has lead to discrepant results in empirical
analyses o f its relationship to memory performance. Previous work examining the
metamemory-memory connection provides little systematic guidance for hypothesis
generation, because the operationalizations o f metamemory appear to play a role in the
results. In a study encompassing the adult life span. Bruce et al. (1982) found that when
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
25
metamemory was operationalized as either the ability to assess memory task demands, or
as prediction o f recall prior to exposure to the memory task, older adults performed
significantly worse than younger adults. However, when metamemory was
operationalized as recall prediction accuracy after the learning tasks, no age differences
were found. This latter finding confirmed some o f the results from an earlier study by
Perlmutter (1978) who found no age differences in metamemory as indicated by
metamemorial knowledge, memory monitoring, memory prediction, and memory
confidence rating, but found significant age differences favoring younger adults in the
spontaneous use o f effective organization strategies. As depicted in Table 2. the most
common ways that studies have measured self-reported memory ability include: open
ended questions: a single "yes" or "no" question: and Likert-scaled questionnaires.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26
Table 2. Measures o f Self-reported Memory Ability in Empirical Studies on the
Relationship Between Self-Reported and Objective Memory Ability.
Study Measure of Self-Reported Memory Ability Outcome
Zelinski et al. (1980) Metamemory Questionnaire (92 items) +
Reige(1982) Memory Self-Report Questionnaire (20 items)
j-
Dixon & Hultsch (1983) Metamemory In Adulthood (120 items) +
Zelinski et al. (1990) #1 Memory Functioning Questionnaire (64 items) +
Zelinski etal. (1990) #2 Memory Functioning Questionnaire (64 items) +
Christensen (1991) Untested Questionnaire (5 items) -r
Bassett & Folstein( 1993) Yes/ No Question: Trouble with Memory? 4-
Fingerman et al. (1994) Rate Memory After a Task
_ i _
Williams et al. (1987) Memory Problem Questionnaire (25 items) +
Zarit (1982) Memory Complaint Interview (11 items) +
Kahn et al. (1975) Open-ended Question: List memory problems -
West etal. (1984) Adapted Metamemory in Adulthood(62 items) -
Scogin (1985) Adapted Metamemory Questionnaire (9 items) -
Sunderland et al. (1986) Subjective Memory Questionnaire (28 items) -
O'Hara etal. (1986) Adapted Metamemory Quest. (10 items) -
Popkinetal. (1982) Adapted Metamemory Quest. (20 items) -
Bolla & Bleecker(1986) Yes/No Question:Memory Concerns? (1 item) -
Poitrenaud (1989) General Rating (1 item) -
Zondennan et al. (1989) Yes/No: Memory Complaints? (1 item) -
Scogin & Rohling (1989) Memory Functioning Questionnaire (64 items) -
Bolla et al. (1991) Metamemory Questionnaire (92 items) -
Hanninen et al. (1994) Memory Questionnaire (10 items) -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
27
Looking first at open-ended questions to elicit self-reported memory, consider a
study by Kahn. Zarit. Hilbert, and Niederehe (1975), in which subjects were asked to
think o f any memory problems they had experienced and then rate the complaints on a
five-point scale. Using this format, subjects were more likely to report recent memory
problems and omit memory problems which may occur frequently, but not come to mind
at the time. Further, some relevant memory problems may be omitted because a subject
is uncertain what constitutes a memory problem (Zelinski. Gilewski. & Thompson.
1980). Kalin et al.'s finding that self-reported memory is not related to memory
performance may reflect the under reporting o f memory' problems that is associated open-
ended measures of self-reported memory ability.
Yet while open-ended questions provide too much structure, a single yes/ no
question about the existence of memory failures is likely to be too restricted in that it
prohibits people from reporting the most common response, that they sometimes have
memory problems in certain areas. For example. Bassett and Folstein (1993)
operationalize self-reported memory by asking, "Do you fin d that you have (rouble with
your memory?" Answers were coded as either "yes " or "no " which masks the response
that most people would select which is. "sometimes” , or in "certain situations". Further,
it is unclear what "trouble with your memory” means. A respondent could assign any
number of meanings to the terms "trouble" ranging from bothersome lapses to
debilitating shortcomings.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
28
The self-reported memory indices with the most empirical support and widespread
use are Likert scaled questionnaires. The two most well-regarded questionnaires are the
Memory Functioning Questionnaire (MFQ; Gilewski & Zelinski, 1988) and the
Metamemory in Adulthood instrument (MIA; Dixon & Hultsch. 1983). The MFQ
consists o f sixty-four items rated on a seven point scale from which four factor scales are
calculated: Frequency o f Forgetting, Seriousness o f Forgetting, Retrospective
Functioning, and Mnemonics Usage. Studies on the relationship between self-reported
memory ability and objective memory ability using the MFQ report that about 10% of
the variance in recall can be accounted for by using the MFQ, although some show-
stronger relationships between self-reported memory and memory performance (Gilewski
& Zelinski. 1980: Williams et al.. 1987: Zelinski et al.. 1990): than do others (Scogin.
1985: O'Hara. 1986). The MIA was designed to reflect the aspect o f familiar,
ecologically representative metamemory behavior (Dixon & Hultsch. 1983). A series of
validity and reliability tests of various items resulted in an eight-factor. 120 item (Likert-
tvpe) instrument. The 8 factors and related dimensions are: knowledge o f memory
strategies, knowledge of memory tasks and processes, knowledge of own memory
capacities, perception of change, activity supportive of memory, memory and state
anxiety, and memory achievement, motivation, and locus of control in memory abilities.
Evidence on the psychometric properties o f the MIA suggests adequate content validity,
factorial validity, and internal consistency. Using the MIA in sample of 116 adults.
Dixon and Hultsch (1983) found that self-reported memory ability predicts objective
memory performance on test recall at all ages.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
29
E. MEASUREMENT OF OBJECTIVE MEMORY PERFORMANCE
The construct, memory performance refers to the ability to remember. The
majority of research on memory performance and self-reported memory ability,
operationalize memory performance using test scores on laboratory tasks such as:
immediate recall o f word lists, delayed recall of word lists, recall o f text, retrieval o f
general knowledge, repetition of numbers (digit span), and face-name learning. As
depicted in Table 3. the specific combination of laboratory tasks varies from study to
study.
Table 3. Measures o f Memory Performance in Empirical Studies on the Relationship
Between Self-Reported and Objective Memory Ability.
Study Measure of Objective Memory Ability Outcome
Zelinski et al. (1980) List recall, list recognition, prose recall -r
Reige (1982) List recall, prose recall, face-name task +
Dixon & Hultsch (1983) Prose recall 4-
Zelinski et al. (1990) #1 List recall, list recog, delayed recall, prose 4-
Zelinski et al. (1990) #2 List recall, delayed recall, mini-mental
j-
Christensen (1991) Prose recall, digit span, paired associates -r
Bassett & Folstein(1993) Mini-mental
4 ~
Fingerman et al. (1994) List recall
Williams et al. (1987) WAIS -r
Zarit (1982) List recall, face-name task, tip-of-tongue
-U
Kahn et al. (1975) List recall -
West et al. (1984) List recall, digit span -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
30
Study Measure o f Objective Memory Ability Outcome
Scogin (85) List recall, face-name task, digit span -
Sunderland et al. (1986) Paired associates, prose recall, face-name task -
O’Hara et al. (1986) List recall, short-portable mental status exam -
Popkinetal. (1982) List recall, delayed recall. list recog. prose -
Bolla & Bleecker (1986) Digit Span, WAIS -
Poitrenaud (1989) Auditory List recall, digit span -
Zonderman et al. (1989) Neurological exam -
Scogin & Rohling (1989) Stroop memory for color/word test -
Bolla et al. (1991) Paired associates, prose recall -
Hanninen et al. (1994) Mini-mental, paired associates -
Using digit span, visual retention, and recall o f word lists. Scogin (1985)
concluded that self-reported memory (MFQ) did not predict memory performance.
However these results may reflect floor and ceiling effects, individuals difference, and/or
the fact that digit span and recall of word lists do not closely resemble everyday memory
tasks. In an attempt to use a memory performance measure that truly reflects everyday
remembering. Martin (1986) had research subjects evaluate their ability to remember
specific things, including appointments. She then examined attendance at a scheduled
research appointment and found a close correspondence between assessments o f memory
for appointments and showing up at a research appointment. Older subjects reported
being better at keeping appointments than younger subjects and their attendance
substantiated the claim. Despite its ingenuity, it is likely that Martin's measure of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
memory performance also captured interest in participating in research, or even level of
conscientiousness. Thus while memory performance tasks should approximate everyday
memory, care should be taken to ensure control o f other variables.
F. HYPOTHESES
The two questions guiding the research are: (I) Do self-reports o f memory ability
accurately reflect objective memory ability? and (2) Are their systematic differences
between those who accurately assess their memory ability as opposed to those who over
or under-estimate their ability? To address these questions, the corresponding null
hypotheses will be tested: (1) self-reported memory ability is not significantly related to
objective memory' performance: and (2) those who accurately assess their memory ability
are not systematically different than those who under-estimate or over-estimate their
memory ability.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
32
CHAPTER III: RESEARCH DESIGN
A. SAMPLES: COMBINING BREADTH AND DEPTH
The data used in this study comes from two sources: the Study of Asset and
Health Dynamics Among the Oldest Old (AHEAD), and Long Beach Longitudinal Study
(LBLS). The LBLS study is a cognitive study of mental abilities and aging while the
AHEAD is an epidemiological survey study aimed at understanding the dynamics of
health and health-related behaviors o f older adults. The first section o f this chapter
examines the content and characteristics o f the AHEAD and LBLS data sets. This is
followed by a detailed review of the measures used in the subsequent analyses. The final
section reviews the analysis procedures that will be used to test each o f the hypotheses.
The Study o f Asset and Health Dynamics Among the Oldest Old
The AHEAD began as an auxiliary survey of the Health and Retirement Survey
(HRS). It was conducted by the University o f Michigan's Survey Research Center
between October 1993 and July 1994. The aim of the AHEAD study is to understand the
impacts and interrelationships of change and transition for older Americans in three major
areas: health, finances, and family structure. Because the questions included in the
interview were designed to reflect the analytic and policy interests o f wide variety of
disciplines, measurement inclusion was highly restricted to maintain a reasonable survey
length.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The AHEAD sample consists o f men and women age 70 and older who were non
institutionalized at the time of the interview. Individuals ages 70 -79 were identified
through the HRS screening of an area probability sample o f households. Individuals age
80 and over were identified using a dual sampling frame: half were identified through the
HRS screen: and the other half from a list frame provided by the Health Care Financing
Administration's (HCFA) Master Enrollment File. Based on the HRS screen and the
HCFA list. 10.229 individuals aged 70 to 104 were selected for interview. O f these. 8223
men and women participated resulting in a response rate of 80% (including those who
used proxies).
For those drawn from the HRS screen, if more than one age-eligible individual
person was living in a household, one person was randomly selected. If the selected
individual was married, an interview was sought with the spouse (regardless o f his or her
age) to provide additional information about the household of the selected individual.
For those drawn from the HCFA frame, if the spouse was also cohort-eligible, the spouse
was part o f the sample in his or her own right, but if the spouse was under age 70. the
interv iew was conducted only to provide additional information. Mexican-American
Hispanics. African-Americans, and households in the state o f Florida, were over sampled
to represent the population.
The complex sampling design requires compensatory weighting for the analyses
of the survey data. Beyond simple compensation for unequal selection probabilities,
factors were used to adjust for geographic and race group differences in response rates, as
well as for the subsampling of households in a small number of dangerous areas. Post
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
34
stratification adjustments were made at the household and person level in order to match
sample demographic distributions with known 1990 Census totals. The household
analysis weight is used for household-level variables and is a composite of five factors:
the housing unit selection weight; an adjustment factor for non-listed segments; an
adjustment factor for subsampled areas; a household nonresponse factor: and a household
post stratification factor. The respondent analysis weight is the product o f the household
analysis weight and the person-level post stratification weight. Age-ineligible
respondents have a value o f zero for the person level weight. All of the analyses used in
this research will be conducted using the respondent-level analysis weight.
A major constraint of the AHEAD study was the need to keep the interview
burden reasonable, especially for the respondents in the oldest age range. With this in
mind, the principal investigators limited the survey protocol to 60 minutes. The survey
section containing the health and cognitive items took 27 minutes, the economic section
28 minutes and the family structure and transfers section took 10 minutes. The survey
was conducted by 130 field interviewers using computer assisted interviewing in either an
over-the-phone or face-to-face session. Most of the interviews with individuals age 80
and were conducted face-to-face in their homes, while subjects under 80 were generally
tested over the telephone. Interviewers were able to arrange face-to-face interviews with
anyone having difficulty doing a phone interview or anyone who preferred a face-to-face
interview.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LBLS Study o f Cognition and Aging
The University o f Southern California/Family Health Plan (LBLS) Study of
Cognition and Aging is a research study designed to assess changes in memory and
intellectual abilities in adults. The LBLS sample includes adults o f all ages, but only
those age 70 and over were included in order to make the LBLS data set more
comparable to the AHEAD data set. Complete data was available for 230 subjects aged
70-95. o f these 52 are longitudinal subjects tested in 1978 and in 1994. and 178 cross-
sectional subjects tested only in 1994.
All o f the subjects were selected from the membership rosters o f the Family
Health Plan (FHP). a health maintenance organization based in Southern California. In
1978. 3000 FHP members ages of 28-36 and 55-87 were solicited to participate in the
study. O f these. 582 agreed to participate for a response rate o f 18%. Although a
response rate of 18% appears to be rather low. it must be noted that participation required
a three hour testing of cognitive abilities, which is a considerable time commitment for
the average individual. Although the sampling is clearly biased by the selective
volunteerism. it remains more representative of the Southern California population than
most studies of cognitive aging whose samples consist almost exclusively of college
graduates in both young and older groups. Based on the 1970 Census, the ethnic
composition reflects the composition of the 65+ population o f Long Beach and Orange
County, consisting of mostly white older adults, with small percentages of African
American. Asian, and Hispanic individuals.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
36
In 1994. a new age matched cross-sectional sample was tested to estimate the
effects o f attrition in the original sample and to enable the perpetuation o f the research.
The cross-sectional sample was again drawn from the FHP membership list. Here the
response rate was even lower at 12%. The lower response rate may reflect the fact that
the testing protocol was longer in 1994 than in the original 1978 testing. As a result of
the extensive time commitment and subject matter that can be perceived as somewhat
threatening, the sample is a highly self-selected. The mean education o f the LBLS sample
was a little over 13 years whereas the mean education in the 70+ population nationwide
is approximately 11 years.
In 1994. participants completed a battery o f memory and intelligence tests in two
2-3 hour sessions. While most sessions were conducted in groups sessions at FHP
facilities, the participation of a significant number o f frail and homebound adults
necessitated a large number of private sessions in homes. In addition to the testing
sessions, subjects also were asked to complete a mail-in packet of questionnaires
consisting o f a personality measure, self-reported memory measure, health behaviors
probe, and personal data inventory.
B. MEASURES
The long testing sessions and extensive take home material used in the LBLS
study allowed for more extensive measurement o f self-reported memory ability and
objective memory performance than the time-restricted AHEAD study. Fortunately, in
addition to the more extensive measures in the LBLS study, there are also abbreviated
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
37
measures that are similar (though not identical) to the measures used in the AHEAD
survey. Because there are roughly comparable variables, several of the same models
could be run on both data sets. This section presents the measures of self-reported
memory ability, objective memory ability, sociodemographic characteristics, self-
reported health indicators, and depression that are used in the AHEAD survey and the
LBLS testing protocol, as well as the more extensive measures of self-reported memorv
ability and objective memory ability used in the LBLS data set only. Following the
review. Table 4 offers a summary o f the measures in both data sets.
Self-Reported Memory Ability
In the AHEAD survey, self-reported memory ability was measured by asking
participants to: (1) rate their memory ability at the present time on a 5 point scale with 5
being Excellent-, and (2) to rate their memory ability as Better, the Same, or Worse than it
was 1 year ago. For those who were selected to participate in the AHEAD study but were
either unwilling, or unable, to complete the survey interview, proxy respondents
answered the questions on their behalf. Proxy respondents were asked to rate the
memory ability o f the selected participant on a 5 point scale with 5 being Excellent and to
rate the subjects memory today as compared to his memory 2 years ago. While the
general measure of self-reported memory ability is the same for self and proxy
respondents, the second measure, comparing change over time, is different in that self
respondents are asked about change over 1 year, and proxy respondents are asked about
change over 2 years. Further, self-respondents are given 3 rating options. {Better. Same.
Worse) while proxy respondents were only offered 2 options {Belter. Worse).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
38
The two questions on self-reported memory ability in the AHEAD survey are
similar to two o f the questions on the Memory Functioning Questionnaire (MFQ) that
subjects completed as part o f the LBLS study. Here, subjects were asked to rate their
memory ability in terms o f the kinds of problems they have on a 7 point scale with 7
being . 'V o Problems and 1 being Major Problems. Participants were also asked to rate
their memory abilities as compared to one year ago on a 7 point scale that ranged from
Much Better to Much Worse.
In addition to these two general questions, the MFQ also queries respondents on
62 other items. From the items, four unit-weight factor scales are calculated: Frequency
o f Forgetting, Seriousness o f Forgetting, Retrospective Functioning, and Mnemonics
Usage. Frequency of Forgetting consists o f 33 questions that ask subjects to rate their
memory in terms of the kinds o f problems they have. Areas covered include: general
problems, problems with names, problems with phone numbers you have just checked,
problems with words, and problems while reading. Retrospective Functioning consists
o f 5 questions that ask respondents to rate their memory abilities compared to the way it
20 years ago. 10 years ago. 5 years ago, 1 year ago. and when they were 18. Seriousness
o f Forgetting consists of 18 questions that ask respondents to rate how serious o f a
problem they consider particular memory failure to be when they forget in various
situations. These situations include forgetting names; phone numbers you just checked,
words, and what you are reading. Mnemonics Usage is made up of 8 questions that ask
people to rate how often they use various techniques to remind themselves o f things.
Such techniques include: Keeping an appointment book, writing yourself reminder notes.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39
make lists o f things to do, making grocery lists, planning schedules in advance, mental
repetition, association with other things, keeping things you in prominent places.
Analyses o f the internal consistency o f the MFQ factors report Cromback's alphas
ranging from .82 to .93 (Gilewski. Zelinski. & Schaie. 1990).
Memory Performance
Both data sets contain a list recall measure o f memory performance. In the
AHEAD study, subjects were read a list of 10 one or two syllable English nouns. After
all 10 words were presented, subjects were given up to 2 minutes to orally recall as many
words as possible. In the LBLS study, subjects studied a typed list of 20 concrete one or
two syllable English nouns for 3.5 minutes and were asked to free recall the word list by-
writing down the words. In order to minimize differences caused by age related slowing,
there was no time limit for recall. Immediate recall was scored as the proportion of
correctly recalled items.
In addition to the free recall tasks. 2 measures of prose recall are included from
the LBLS. In the prose recall tasks, subjects read a 227 word essay while listening to the
story7 being read at approximately 155 words per minute. Subjects were asked to recall
the story and write down what they remembered. Subjects were encouraged to recall the
text verbatim but also to include anything that they remembered in their own words.
Prose recall was scored by parsing the passage into content units and relations between
the contents units with Meyer's (1975) system o f prose analysis. The score consists of
the proportion o f content units and relations between units correctly recalled. The more
extensive measures of self-reported memory will be used to examine measurement.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
Self-Reported Health Characteristics
Self-reported indicators o f health include: a self-rating of general health status, a
self-rating of hearing ability, and a self-rating of eyesight. In the AHEAD survey
respondents were asked to rate their health on a 5 point scale with 5 being Excellent and 1
being Very Poor. In the LBLS study, respondents were asked to rate their health.
compared to people their age, on a 6 point scale with 6 Very Good and 1 being Very-
Poor. In the literature on self-reported health, the research suggests that the inclusion o f
the phrase "compared to people your age" does not make answers significantly different
than the exclusion of the phrase because adults tend to compare themselves to people
their own age even when not explicitly asked to do so. In the AHEAD survey, self-rated
eyesight is measured by asking respondents to rate their eyesight with their glasses on a 5
point scale with 5 being Excellent and 1 being Poor. Similarly, respondents rate their
hearing on the same scale with their hearing aid on. This differs from the LBLS measure
of self-rated eyesight and self-rated hearing because here the subjects are asked to rate
eyesight/hearing, compared to people their age. with no mention of glasses or hearing
aids. This difference will be important to keep in mind because a person who has poor
vision that is adequately corrected with eyeglasses would most likely rate her eyesight as
Moderate or Good if asked the AHEAD phrasing of the question but Poor if asked the
LBLS question. Despite the differences, the measures of sensory functioning are still of
interest as they both give us information about how people self-rate their abilities. It will
be interesting to see if respondents who give themselves more negative self-ratings of
memory ability will also give themselves negative self-rating o f the health indicators.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
41
Depression
In the AHEAD study, depression was measured using eight items of the Clinical
Epidemiological Survey of Depression (CESD). In the reduced CESD. four items reflect
mood, and three items reflect the psychosomatic dimension o f depression. The response
scale was simplified into two categories (yes & no). Kouhout et al. (1993) demonstrated
that these modification sacrifice little of the structure and precision of the original CESD
scale. In the LBLS study, depression was measured using the Geriatric Depression Scale
(GDS). The GDS consists of thirty yes/no questions. Because the physiological changes
associated with aging can be similar to those associated with depression, the GDS was
specifically designed to exclude somatic symptomology o f depression. The GDS was
completed at the beginning of the testing session. Table 4 summarizes the measures used
in the two data sets. The measures that are comparable in both data sets (the abbreviated
measures) are placed across from each other in the columns while the more extensive
measures and the measure of neuroticism. are listed only in the LBLS column.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
42
Table 4. A Comparison o f the Measures Used in the LBLS and AHEAD Data Sets.
VARIABLE LBLS DATA
N = 230
AHEAD DATA
N = 6436
Self-Reported
Memory Ability
Self-rating of Memory
How would you rate your
memory in terms of the kinds
o f problems that you have?
No Problems = 7
= 6
= 5
Some Minor Problems = 4
= 3
= 2
Major Problems = 1
Self-rating 1 year ago
How is your memory
compared to the way it was
one year ago?
Much Better = 7
Better = 6
Somewhat Better = 5
Same = 4
Somewhat Worse = 3
Worse = 2
Much Worse = 1
Self-Reported
Frequency of Forgetting
General Rating
Names/Faces
Appointments
Where you put things
Performing household chores
Directions to places
Phone no. you use freq or just checked
Things people tell you
Keeping up correspondence
Personal dates
Forget what you wanted to buy at store
Begin something forget what doing
Losing thread in coversation
Self-rating o f Memory
How would you rate your
memory at the present time?
Excellent = 5
Very Good = 4
Good = 3
Fair = 2
Poor = 1
Self-rating 1 year ago
Compared with 1 year ago.
would you say your memory is
better, the same or worse now?
Better = 3
Same = 2
Worse = I
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
43
VARIABLE LBLS DATA
N = 230
AHEAD DATA
N = 6436
Self-Reported
Seriousness of Forgetting
General Rating
Names/Faces
Appointments
Where you put things
Performing household chores
Directions to places
Phone numbers you just checked
Phone numbers used frequently
Things people tell you
Keeping up correspondence
Personal dates
Words
Forget what you wanted to buy at store
Taking a test
Begin something forget what doing
Losing thread in conversation
Knowing already told someone
Self-Reported
Retrospective Functioning
Memory compared to:
One year ago
Five years ago
Ten years ago
Twenty years ago
W hen you were eighteen
Self-Reported
Mnemonics Usage
Keep an appointment book
Write yourself reminder notes
Make lists of things to do
Make grocery list
Plan daily schedule in advance
Mental repetition
Associations with other things
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
44
VARIABLE LBLS DATA
N = 230
AHEAD DATA
N = 6436
Objective Memory
Performance
Immediate Recall
of 20 word list
hom o
H ag
bird
dirt
woman
ocoan
church
iron
vest
exam
rattle
money
jury
lemon
star
alcohol
kettle
garden
painter
tank
Prose Recall
"Parakeets are ideal pets for people with
limited time space and money..."
Immediate Recall
of 10 word list
arm y
bird
car
door
frost
lake
mountain
plant
ticket
winter
Sociodemographic
Variables
Sex
Male = 1
Female = 2
Sex
Male = 1
Female = 2
Chronological Age
Ages: 70-95
Chronological Age
Ages: 70-103
Education
Number of Years: 6-17+
Education
Number o f Years: 0-17+
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
45
VARIABLE LBLS DATA
N = 230
AHEAD DATA
N = 6436
Self-Reported
Health
Indicators
Self-Rated Health
Compared to other people my
age, I believe my health to be:
Very Good = 6
Good = 5
Moderately Good = 4
Moderately Poor = 3
Poor = 2
Very Poor = 1
Self-Rated Eyesight
Compared to other people my
age, I believe my eyesight to
be:
Very Good = 6
Good = 5
Moderately Good = 4
Moderately Poor = 3
Poor = 2
Very Poor = I
Self-Rated Hearing
Compared to other people my
age, I believe my hearing to
be:
Very Good = 6
Good =5
Moderately Good = 4
Moderately Poor = 3
Poor =2
V ery' Poor = 1
Self-Rated Health
Would you say that your
health is :
Excellent = 5
Very Good = 4
Good = 3
Fair = 2
Poor = 1
Self-Rated Eyesight
(With your glasses on), is your
eyesight:
Excellent = 5
Very Good = 4
Good = 3
Fair = 2
Poor = 1
Self-Rated Hearing
(With your hearing aid on) is
your hearing:
Excellent = 5
V ery' Good = 4
Good = 3
Fair = 2
Poor = 1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
46
VARIABLE LBLS DATA
N = 230
AHEAD DATA
N = 6436
Depression
Depressive Symptoms
30 ves/no items:
Basically satisfied
Dropped Activities
Life empty
Often bored
Hopeful about future
Thoughts in head
Good spirits
Afraid something bad
Happy
Restless and fidgety
Prefer stay home
Worry about future
More problems than most
Wonderful to be alive
Downhearted and blue
Your pretty worthless
Worry about past
Find life exciting
Hard to start project
Full of energy
Your situation hopeless
Most better off
Upset
Crying
Trouble concentrating
Enjoy getting up
Avoid gatherings
Easy to make decision
Mind clear
Depressive Symptoms
8 ves/no items:
Felt depressed
Everything w as an effort
Sleep restless
Happy
Lonely
Enjoyed life
Felt sad
Could not get going
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
47
C. ANALYSIS PROCEDURES
Descriptivcs
Before testing the hypotheses, several descriptive analyses were conducted. First,
the composition o f the two samples were conducting a series of cross- tabulations to
compare the age. sex. education, health (self-reported), and level of depression o f the two
samples. Next, the mean scores on all the comparable measures of the two data sets were
compared, making special note o f scaling differences. Third, correlation matrices were
constructed to compare the level o f association between variables in the two data sets.
Hypothesis One: Self-reported memory ability is not significantly related to
objective memory performance.
In order to determine whether the general self-rating of memory ability was
related to memory performance even when the effects of other variables thought to
impact memory ability had been partialled out, hierarchical regression analyses were
performed, with predictor variables entered in blocks. This method, which involves
forced entry into the regression equation, makes it possible to test the effects of a set of
variables when the effects o f another set, believed to be correlated with that set. have
already been partialled out and tested for level of association. Two competing
hierarchical regression models were calculated to determine the relative contribution of
the predictor variables. In Model I, the sociodemographic characteristics, age. sex. and
education were entered in the first predictor block. Being older, male, and having less
education are associated with worse performance on list recall memory tasks so our aim
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
48
was to partial out the effects of these variables. Next, the health indicators, self-rated
health, self-rated hearing, and self-rated eyesight were entered in the second block. More
negative self-ratings o f health, eyesight and hearing have found to predict worse health,
eyesight, and hearing which in turn have been shown to be associated with worse
performance on memory tasks. Depression was entered in the third block as a significant
association has been shown to exist between depression and memory impairment. The
predictor variable, general self-rated memory ability was entered into the fourth block. In
Model 2. the two measures of self-reported memory ability were entered in the first
block, depression in the second, self-rated indicators of health in the third, and the
demographic characteristics were entered last. Within each o f the predictor blocks, the
variables were variously ordered to examine their explanatory contribution and possible
interaction with other independent variables in the model.
Next, to examine whether data obtained by proxy would differ data obtained by
self-respondents. The characteristics o f those who had proxy respondents were compared
to the characteristics o f the rest o f the sample. Next the proportion of proxy respondents
who selected each rating to the proportion o f self-respondents who selected each rating
were compared. Because those with proxy respondents are fundamentally different (age.
sex. education, health) than the self-respondents, it is impossible to say whether the
differences in the ratings are the result of proxy response. If proxy respondents rate
memory ability in a way that is systematically different from those who rate their own
memory abilities, then it is inappropriate to use proxy ratings to indicate memory ability.
Work by Zelinski et al. (1983) suggests that proxy ratings of memory are not
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
systematically different from self-ratings o f memory ability, but this proposition cannot
be tested here . It was. however, possible to examine whether adult children would rate
memory ability more negatively than spouses by comparing the mean ratings o f the two
proxy relationship groups using a t-test of statistical significance.
In the next set o f analyses, the 64 item Memory Functioning Questionnaire from
the LBLS study was used to examine the role o f measurement of self-reported memory
ability. In addition to the general rating of memory ability and rating o f memory ability-
compared to one year ago, the MFQ queries respondents on 62 other items that comprise
4 unit-weight factor scales: Frequency o f Forgetting, Seriousness o f Forgetting,
Retrospective Functioning, and Mnemonics Usage.
The item that is most similar to the single item self-rating of memory is the
factor. Frequency o f Forgetting. The item that is closest to the single item rating o f
memory ability 1 year ago is the factor, Retrospective Functioning. There was no
equivalent abbreviated measure for the factors Seriousness of Forgetting or Mnemonics
Usage.
The equations previously run on data from both data sets, were compared to the
model run with more extensive measures of self-reported memory ability. The analyses
were only conducted on the LBLS sample because the AHEAD data set did not include
the more comprehensive measure of self-reported memory ability. The MFQ factor.
Mnemonics Usage, was not included in the final analyses because it did not account for
any additional variance, so using the three factors only resulted in a more parsimonious
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
50
model. As with the analyses of the abbreviated measures, two models were used to
examine the relative contribution of the predictor variables to memory performance.
Just as the adequacy o f the measurement o f self-reported memory may affect the
appearance o f its relationship with objective memory performance, the adequacy o f the
measure o f objective memory performance may also impact the appearance o f the
relationship. The LBLS study includes two measures o f prose recall. In the prose recall
tasks, subjects read a 227 word essay, while listening to it aloud. Participants are
instructed to write down anything they can remember from the essay. Interviewers
encourage participants to recall the text verbatim if possible, or else to include anything
that they remember in their own words. Prose recall is scored by parsing the passage into
content units and relations between the content units. The score consists o f the
proportion o f correctly recalled content and relation units. Because only the LBLS
contains more extensive measures of objective memory performance, these analyses
could only be conducted upon the LBLS sample.
Hypothesis Two: Those who accurately appraise their memory ability are not
systematically different from those who over-estimate or under-estimate their
memory ability.
First, in order to examine the accuracy o f self-ratings of memory ability, it was
necessary to recode self-reported memory abiltiy and objective memory performance into
three categories. The five-point scale for the general rating of memory item on the
AHEAD was reduced to a three-point scale by combining the top two ratings into a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
51
category labled "good”; the middle rating into “average”; and the lowest two ratings into
"poor". In the LBLS data set, the top three rankings were combined into the category
labeled “good”; the middle two ratings into “average”; and the lowest three ratings were
combined into a category labeled “poor”.
Cross-tabulations were then conducted in order to determine what percentage o f
respondents were accurate in their assessment; what percentage under-estimated memory
ability: and what percentage over-estimated memory ability. To determine whether there
were systematic differences between the participants who accurately assessed their
memory ability and the participants who did not, a multinomial logistic regression
analysis was conducted on the AHEAD data set. The small size o f the LBLS sample
precluded the use o f the LBLS data in the logistic analyses. Unlike ordinary least squares
regression, multinomial logistic regression is able to model the relationship between a
three category dependent variable and categorical and continuous independent variables.
The result of the equation is a model of the probability o f occurrence of the dependent
variable, controlling for known explanatory variables. In this analysis, the categories o f
the dependent variables were under-estimation o f memory ability, over-estimation o f
memory ability. and accurate estimation o f memory ability. Accurate estimation was
omitted as the comparison category. Because accurate estimation was omitted, the results
reflect the probability o f over-estimation (or under-estimation) relative to accurate
estimation of memory ability controlling for known explanatory variables.
The results of the logit model are displayed showing the parameter estimates,
marginal probabilities, and odds ratios. The parameter estimates estimate the extent to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
52
which each independent variables explain the relevant category of the dependent variable.
The marginal probability of each independent variable is calculated to determine how
much impact each independent variable has on the relevent dependent variable. For
continuous independent variables, the marginal probability is calculated using the
formula. Pe>(X p). where a is the normal density function and X is the individual's value
of the independent varaible. For dummy independent variables, marginal probabilities are
the sample mean o f the difference between the predicted probability o f scoring one rather
than zero in the independent variable o f interest. Odds ratios are calculated by
exponentiating the parameter estimate for each explanatory variable. The resulting figure
represents the ratio odds of a one-unit change in the independent variable. The odds ratio
gives an estimate o f the relative effect o f the independent variable in comparison to the
omitted values.
Using the results of the multinomial logistic regression, a profile o f the
characteristics associated with under-estimating and over-estimating memory ability was
constructed. To determine the relative contribution of each of the characteristics in the
profile, a series of vignettes were constructed in order to see the change in probability
associated with manipulating an individual characteristic. From the results o f the vignette
analysis, the characteristics associated with over- and under-estimating memory ability
were ranked in order o f importance in order to provide a more accurate profile.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER IV: THE RELATIONSHIP BETWEEN SELF-REPORTED
AND OBJECTIVE MEMORY ABILITY
A. DESCRIPTIVES
O f the 8.223 participants interviewed for the AHEAD study. 776 were age
ineligible (under age 70). leaving a sample o f 7,447. O f these. 768 had proxy
respondents. leaving a self-respondent sample o f 6,679. Data was missing in 243 o f these
subjects leaving a final sample of 6.436 self respondents and 768 proxy respondents.
The descriptive analyses, and all o f the subsequent analyses conducted on the AHEAD
sample, used the respondent level weight to compensate for unequal selection
probabilities and to adjust for geographic and race group differences in response rates.
In the LBLS study. 428 people over the age o f 70 participated in the study, but
complete data was available for 230. In over 75% o f the cases, incomplete data was the
result of the subject not mailing back the “take home” questionnaires that covered the
self-rated indicators of health and self-rated memory ability.
A notable difference between the two samples was their sex distributions
(Table 5). The AHEAD sample is comprised o f 37% men and 63% women, while the
LBLS sample is made up of 53% men and 47% women. Matched with known 1990
Census totals, the AHEAD reflects the gender distribution of the 70+ population which
has a far greater number of women than men as a result of mortality differentials
between the sexes. The sex distribution o f the LBLS sample does not reflect the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
54
population and instead reflects an effort on the part o f the researchers to recruit
enoughmen to attain an approximately equal number of women and men.
Table 5. Sex Distribution of Respondents in the AHEAD and LBLS Samples
AHEAD n=6436 LBLS n=230
Number Percent Number Percent
Men 2381 37% 123 53%
Women 4055 63% 107 47%
The average age o f participants was 77.06 years (SD=5.53) in the AHEAD and
78.99 (SD= 5.70) in the LBLS. As depicted in Table 6, The AHEAD sample has a
greater proportion o f younger participants than does the LBLS. Forty percent o f the
AHEAD sample is between the ages 70 and 74 as compared to only 26% o f the LBLS
sample. In the other age groups, the two data sets are more comparable with a decreasing
proportion o f subjects in the older age groups. In the AHEAD, the distribution o f men
and women in each age group approximates the overall age distribution, with men
accounting for approximately a third of each of the age groups. In the LBLS sample,
there are more men than women ages 70 and 84, but at ages 85 and older, men account
for approximately a third of each age group.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
Table 6. Age Distribution of Respondents in the AHEAD and LBLS Samples
AHEAD (n=6436) LBLS (n=230)
Percent Percent
70-74 40% 26%
75-79 30% 27%
80-84 19% 27%
85-89 8% 15%
90-94 2% 4%
95-99 <1% <1%
100+ <1% <1%
Total 100% 100%
The racial composition of the two samples also differed. Ten percent of the
AHEAD sample were African American, as opposed to only 2% of the LBLS sample. In
both samples, participants could identify themselves as being White and Hispanic or
African American and Hispanic. Four percent o f the AHEAD sample was Hispanic and
3% of the LBLS sample.
Table 7. Racial and Ethnic Composition of the AHEAD and LBLS Samples.
AHEAD (n=6436) LBLS (n=230)
White 90% 99%
African American 10% 2%
Hispanic 4% 3%
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
56
In the AHEAD sample, the mean number of years o f education is 11.18, with a
standard deviation o f 3.50 years, while the LBLS sample had a mean of 13.28 years with
a standard deviation o f 3.04 years. The difference in education between the two samples
is likely the result o f a rigorous testing protocol and subsequent self-selection o f more
educated people in the LBLS sample. As depicted in Table 7. 9% of the AHEAD sample
did not have any education beyond grade school as compared to only 1% of the LBLS
sample. The majority of the AHEAD sample completed up to high-school. while the
majority of subjects in the LBLS sample attended college. In both samples, the education
distribution between men and women was comparable, except for graduate school, where
men out numbered women.
Table 8. Years of Education in the AHEAD and LBLS Samples.
AHEAD LBLS
G rade School: 0-6 years 9% <1%
Junior High: 7-8 years 14% 7%
High-school: 9-12 years 48% 24 %
College: 13-17 years 22% 36%
G raduate: 17+ years 6% 14%
While all of the LBLS sample resided in California, the AHEAD sample was
selected from States across the nation. Reflecting the U.S. population of adults over age
70. 16% of the sample lived in the Northern region o f the United States. 32% in the
Southern region. 40% in the Eastern region, and 16% in the Western region (Table 9).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
57
Table 9. U.S. Region of Residence.
AHEAD (n=6436) LBLS (n=230)
North 16% 0
South 32% 0
East 40% 0
West 16% 100%
In both samples health status was measured using a self-reported rating o f health
on a Likert-type scale. The AHEAD survey uses a 5 point rating scale with choices
ranging from Excellent to Poor. while the LBLS survey uses a 6 point rating scale with
choices ranging from Very Good to Very Poor. Exact comparisons cannot be made
because o f the differences between the scales, but it is possible to examine the proportion
of respondents who selected each response. As depicted in Table 10 and Table 11.11%
o f the AHEAD sample selected the highest health rating as compared to 28% o f the
LBLS sample. While this difference may reflect differences in the wording o f the
highest rating option (Excellent in the AHEAD and Very Good in the LBLS). it is also
likely to reflect actual differences in the health of the samples. The demanding nature of
the LBLS protocol is likely to have biased the sample toward those in better health. Ten
percent o f the AHEAD population rated their healdi in the lowest ranking {Poor), as
opposed to 1% in the last two categories o f the LBLS sample {Poor and Very Poor), and
only 7% in the last three categories {Moderately Poor, Poor, and Very Poor). Within
each o f the data sets, men and women were similar in their health ratings. The proportion
of men and women in each of the response ratings was comparable. For example, in the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
58
AHEAD 13% o f the men, and 10 % of the women in the sample gave themselves the
highest rating, and in the LBLS sample 28% o f the men, and 29% of the women gave
themselves the highest rating.
Table 10. Self-Reported Health by Sex in the AHEAD Data Set.
AHEAD n=6436
Health Rating Men Women Total
Excellent: 5 13% 10% 11%
Very Good: 4 23 25 24
Good: 3
•> n
j j 31 32
F air: 2 22 23 22
Poor: 1 10 1 1 10
Table 11. Self-Reported Health by Sex in the LBLS Data Set.
LBLS n=230
Health Rating Men Women Total
Very Good: 6 28% 29% 28%
Good: 5 35 34 34
Moderately Good: 4 30 30 30
Moderately P o o r: 3 7 6 6
Poor: 2 1
•*>
J
Very Poor: 1 — 1 .4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
59
To determine whether differences in self-rated health status in the two data sets
were a function o f age differences between the two data sets, cross tabulations were
conducted on self-reported health and age group. Interestingly, there was little variation
in self-ratings o f health among the age groups. In the AHEAD sample (Table 12). the
proportion of subjects who selected the highest rating is near 10% in each o f the age
groups, as opposed to the LBLS sample (Table 13), where the proportion o f subjects who
selected the highest rating is near 30% in all of the age groups.
Table 12. Self-Reported Health by Age Groups in the AHEAD Data Set.
AHEAD n=6436
70-74 75-79 80-84 85-89 90-94 95+
Excellent: 5 13% 11% 10% 9% 9% 6%
Very Good: 4 27 22 21 23 20 31
Good: 3 32 32 31 J J j j 34
Fair: 2 20 25 24 22 22 16
Poor: 1 8 10 14 12 16 13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 13. Self-Reported Health by Age Groups in the LBLS Data Set.
60
LBLS n=230
70-74 75-79 80-84 85-89 90-94 95+
Very Good: 6 27% 33% 19% 31% 44% 100%
Good: 5 39 34 34 29 j j 0
Moderately Good: 4 24 23 42 31 22 0
Moderately Poor: 3 5 8 5 9 0 0
Poor: 2
* *
2 0 0 0 0
Very Poor: 1
• >
2 0 0 0 0
In addition to self-reported health status, both data sets contain measures o f self-
reported eyesight and self-reported hearing. Like the self-reported health questions, there
are differences between the two data sets in the way the sensory questions are posed. In
the AHEAD survey, self-rated eyesight is measured by asking respondents to rate their
eyesight with their glasses on, on a 5 point scale, and their hearing on the same scale.
with their hearing aid on. This difference is likely to be important in that a respondent
who has poor vision that is adequately corrected with eyeglasses, would most likely rate
her eyesight as Moderate or Good, if asked the AHEAD question, but Poor if asked the
LBLS question.
In the AHEAD sample, the mean self-rating of eyesight was 3.20 on a 5 point
scale with a standard deviation o f 1.09. In the LBLS sample, the mean was 4.34 on a 6
point scale with a standard deviation of 1.11. Table 14 and Table 15 summarize the
scores on self-reported eyesight. The distribution of responses is similar between the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
61
two data sets with 12% o f respondents in the AHEAD sample selecting the highest rating
o f eyesight, and 13% o f the LBLS sample. The two most frequently selected responses
were the second and third highest responses. Sixty-four percent o f the AHEAD sample
selected the ratings (Very Good and Good), as did 71% o f the LBLS sample (Good and
Moderately Good). In the AHEAD sample, the proportion o f men and women who
selected each response was very similar. In contrast, the men and women in the LBLS
tended to select different ratings. Twenty-one percent of the women rated their health as
Excellent as opposed to 6% o f the men. Instead, men were more inclined to rate their
health as Good (41%) as opposed to women (24%).
Table 14. Self-Reported Eyesight by Sex in the AHEAD Data Set.
AHEAD n=6436
Men Women Total
Excellent: 5 15% 11% 12%
Very Good: 4 27 27 27
Good: 3 36 37 37
F air: 2 16 17 17
Poor: 1 6 9 7
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 15. Self-Reported Eyesight by Sex in the LBLS Data Set.
62
LBLS n=230
Men Women Total
Very Good: 6 6% 21% 13%
Good: 5 41 24 34
Moderately Good: 4 40 35 37
Moderately P oor: 3 8 11 10
Poor: 2 2 6 4
Very Poor: 1 2
J
The same descriptive analyses were conducted on self-rated hearing (Table 16 and
Table 17). The mean rating was 3.27. on a 5 point scale, with a standard deviation of
1.07 in the AHEAD, and in the LBLS, the mean was 4.17 on 6 point scale, with a
standard deviation of 1.26. Like self-reported eyesight, the results were comparable
between the two samples. In the LBLS sample, men were less likely than women to give
themselves the highest rating, and more likely to give themselves the second highest
rating.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
63
Table 16. Self-Reported Hearing by Sex in the AHEAD Data Set.
AHEAD n=6436
Men Women Total
Excellent: 5 12% 17% 15%
Very Good: 4 22 26 24
Good: 3 37 37 37
F a ir: 2 24 16 19
Poor: 1 6 4 4
Table 17. Self-Reported Hearing by Sex in the LBLS Data Set.
LBLS n=230
Men Women Total
Very Good: 6 8% 21% 14
Good: 5 23 28
Moderately Good: 4 30 34 32
Moderately P oor: 3 15 13 14
Poor: 2 12 4 8
Very Poor: I 2
5
Turning now from indicators o f health to an indicator of affective status, recall
that the abbreviated version of the Center for Epidemiologic Studies of Depression
(CESD-8) Scale was used to measure depression in the AHEAD. The CESD-8 consists
of 8 items reflecting affective, interpersonal, and psychosomatic symptoms. In contrast.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
64
the Geriatric Depression Scale (GDS) used in the LBLS sample, consists of 30 items
reflecting affective and interpersonal symptoms but not psychosomatic symptoms. The
exclusion o f psychosomatic symptoms on the GDS reflects the concern that
psychosomatic symptoms among the elderly may represent symptoms o f comorbid
physical disease rather than depression. The question then, is whether results o f the two
depression scales can be compared. Studies by Kessler et al. (1992) and Heithoff (1995)
suggest that comparison may be justified since pattern o f findings are maintained when
psychosomatic symptoms are removed from the CESD8-D. However, because there are
only a possible o f 8 depression symptoms on the CESD8-D and a possible o f 30
depressive symptoms on the GDS, it is necessary to group the results in order to make
meaningful comparisons. Herzog (1995) grouped the results o f the abbreviated CESD8-
D into three categories: 0-1 symptoms, 2- 3 symptoms, and 4+ symptoms. This three
tiered grouping system can be compared to three groupings of the GDS: 0-7 symptoms.
8-13 symptoms, and 14-30 symptoms (Yesavage, 1983). Based on these groupings.
Table 18 and Table 19 summarize the depressive symptoms in the two samples. In the
AHEAD sample, 62% of the sample had few depressive symptoms. 22% has some
depressive symptoms, and 16% had enough depressive symptoms to suggest depression
in the AHEAD. The mean number of depressive symptoms was 1.61 with a standard
deviation of 1.95. In the LBLS sample. 80% of the sample had few depressive
symptoms. 14% had some depressive symptoms and 6% had scores indicating
depression. Here, the mean number of depressive symptoms was 4.99. with a standard
deviation of 4.61.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
65
The lower proportion o f respondents reporting depressive symptoms in the LBLS
sample could be a reflection o f the measure of depression used and/or a reflection o f the
population. It may be that those who are depressed are less likely to agree to participate
in a relatively taxing study of cognition. In both samples, women exhibit more
depressive symptoms than do men.
Table 18. Depression Grouping by Sex in the AHEAD Data Set
AHEAD n=654
Men Women Total
Depression unlikely
0-1 Depressive Symptoms
69% 58% 62%
Possible Depression
2-3 Depressive Symptoms
19 23 22
Probable Depression
4+ Depressive Symptoms
12 19 16.
Table 19. Depression Groupings by Sex in the LBLS Data Set.
LBLS n=230
Men Women Total
D epression Unlikely
0-7 D epressive Sym ptom s
84% 77% 80%
P ossible Depression
8-13 D epressive Sym ptom s
15 13 14
P robable Depression
14-30 D epressive Sym ptom s
2
10 6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
66
Having examined the sex, age, education, health status, and affective status of the
two samples, it is possible to consider the relationships between the variables in each
data set. Before presenting the intercorrelation matrices, Table 20 offers a review of the
variables in both data set by presenting the mean scores, standard deviations, and scales
o f each of the measures.
Table 20. Means and Standard Deviation of the Comparable Measures in the AHEAD
and LBLS data sets.
AHEAD N=6436 LBLS n=230
Scale Mean SD Scale Mean SD
Self-reported
Health
1 -5 3.03 1.15 1 -6 4.81 .99
Self-reported
Eyesight
1 -5 3.20 1.09 1 -6 4.34 1.1 1
Self-reported
Hearing
1 -5 3.27 1.07 1 -6 4.17 1.26
Self-Rated
Memory
Ability
1 -5 3.16 .99 1 -7 4.6 1.34
Self-rated
Memory lyr
comparison
1-3 1.55 .JJ 1 -7 3.94 1.01
Memory
Ability on List
Recall
Proportion
correct of
10 words
.46 .19 Proportion
correct of
20 words
.51 .19
To examine the strength o f the relationships between the variables. Pearson
product-moment correlation matrices were constructed for each data set (Table 21 and
Table 22.). Because o f the disparity in sample size, some correlation coefficients of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
67
comparable sizes were significant in the AHEAD sample, but not in the LBLS sample.
For all the variables that were significantly correlated in both data sets, the direction and
approximate magnitude was the same. In both data sets, age was significantly negatively
correlated with education, self-rated hearing, both self-rated memory measures, and
objective memory performance. Sex (being a woman) was negatively correlated with
education, and positively correlated with being depressed, and surprisingly, with a higher
self-rating o f memory ability. Education was significantly correlated with a better health
rating and scoring better on the memory performance task, and negatively correlated with
depression. Self-rated health, eyesight, and hearing were all positively correlated, and
negatively correlated with depression. Self-rated memory ability was positively
correlated with self-rated health and hearing, but not vision, and also positively correlated
with the self-rated memory 1 year comparison, and objective memory performance on the
list recall task.
In the AHEAD data set. the variable relationships that had the highest bivariate
regression coefficients were education and memory performance (.37). and self-reported
health and self-reported vision (.37). In the LBLS sample, the most closely related
variables were self-reported memory ability and self-reported memory ability I year
comparison (.40). and self-reported health and depression (.37). While the coefficients
of these correlations are somewhat higher than most of the other coefficients, their
magnitude does not suggest the presence o f multicollinearity as .60 is the common
standard.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission.
68
Table 21. Intercorrelalions Between Variables in the AHEAD Data set
1
2
3 4 5 6 7 8 9 10 1 1 12 13
1. Age 1
2. Sex .09** 1
3. Black .04* .03* 1
4. Hispanic -.02* .01* -.05** 1
5. Education -.13** -.04* -.26**
. 27**
1
6. Self-Report Health -.10** -.03* -.11** -.08** .27** 1
7. Self-Report Vision -.21** -.06** -.09** -.03* .21** .37** 1
8 Self-Report Hearing -.16** .13** -.01 -.02* .15** .24** .30** 1
9. Depression .12** .11** .07** .11** -.13** -.41 ** -.26** -.17** 1
10. Self-Reported
M emory
-.10** .05** -.05** -.04* .18** .28** .26** .29** -.20** 1
11. Self-Reported
M emory 1 Y ear Ago
-.09** .04* -.02* -.01* .06** .15** .12** .10** -.19** .28** 1
12. Objective Memory -.31** .08** -.20** -.08** .37** .20** .20** .20** -.19** .18** .12** I
13. Southern Region -.001 -.005 .12** .11** -.16** -.07** -.08** -.07** .05** -.06** -.01* -.1* 1
*p<.05, **p<.001
Reproduced with permission o f th e copyright owner. Further reproduction prohibited without permission.
69
Table 22. Inlercorrelations Between Variables in the LBLS Data Set.
1
2
4 5 6 7 8 9 10 11 12
1 • Age 1
2. Sex .09 1
3. African American -.06 -.06 1
4. Hispanic .1 -.1) -.01 1
5. Education -.16* -.17* -.05 -.1 1
6. Self-Report Health .03 .01 .01 -.02 .13* 1
7. Self-Resport Vision -.1 .02 .04 -.08 .04 .31** 1
8. Self-Report Hearing -.16* .11 .04 -.13 .01 .20** .31* 1
9. Depression .12 .18* .00 -.03 -.12* -.37** -.16* -.26** 1
10. Self-Report Memory -.13* -.01 -.03 -.004 .1 .27** .11 .19* -.21* 1
11. Self-Report Memory 1
Year Ago
-.14* -.01 .20 .04 .04 .04 .13 .10 -.13* .40** 1
12. Objective Memory -.30** .17* -.04 -.13 .12* .07 -.01 .10 .03 .16* .08 1
*p<.05, **p<.001
70
B. THE ROLE OF SAMPLING
The AHEAD's 5 point memory self-rating scale ranges from Excellent to Poor.
while the LBLS's 7 point rating scale ranges from No Problems to Major Problems.
Again, while exact comparisons cannot be made because o f die differences between the
scales, it is possible to examine the proportion of respondents who selected each
response. As depicted in Tables 23 and Table 24. approximately 10% o f the respondents
in both samples selected the highest rating, 40% of the respondents selected the middle
option, and less than 5% selected the lowest rating o f memory ability. In both samples
the responses o f men and women were fairly similar. In the AHEAD sample 75% of the
respondents rated their memory ability as being Excellent, Very Good. or Good, which is
contrary to the popular notion that most older adults have complaints about their memory
abilities. Similarly, in the LBLS sample, only 1% o f the sample reported having Major
Problems with memory ability and only 5% picked either o f the two lowest rankings.
Table 23. Self-Reported Memory Ability by Sex in the AHEAD Data Set.
AHEAD n = 6436
Men Women Total
Excellent: 5 9% 9% 9%
Very Good 4 24 27 26
Good: 3 38 41 40
F a ir: 2 23 18 20
Poor: 1 5 4 4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
71
Table 24. Self-Reported Memory Ability by Sex in the LBLS Data Set.
LBLS n = 230
Men Women Total
No Problems: 7 10% 11% 10%
6 15 18 17
5 22 12 17
Some Problems:4 36 45 40
j 14 7 11
2
4 4
Major Problems:! 0
* i
j 1
Next, the memory self-ratings o f various age groups were compared. Table 25 and
Table 26 present the results of cross tabulation analyses of self-reported memory ability
by age group. In the AHEAD sample, the proportion of people who gave themselves the
highest or second highest rating. Excellent or Very Good did not differ by age group. For
the middle rating. Good, 42% of those in the younger age groups (age 70-74 & 75-79)
selected this rating, as compared to 35% in the older age groups. This pattern held in
reverse for the lower ratings of Fair and Poor, with the older groups having a somewhat
larger proportion of respondents than the two groups in their 70s. This pattern was also
observed in the LBLS sample, although the number of subjects in the oldest age groups
prohibits generalizations.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
72
Table 25. Self-Reported Memory Ability by Age Group in the AHEAD Data Set.
AHEAD n=6436
70-74 75-79 80-84 85-89 90-94 95+
Excellent: 5 9% 9% 9% 12% 8% 9%
Very Good: 4 30 25 22 20 20 34
Good: 3 41 43 36 38 35 31
Fair: 2 17 19 26 25 30 19
Poor: 1
2
4 7 5 8 6
Table 26. Self-Reported Memory Ability by Age Group in the LBLS Data Set.
LBLS n=230
70-74 75-79 80-84 85-89 90-94 95+
No Problems: 7 10% 10% 13% 11% 0% 0%
6 23 21 8 1 1 0 100
5 1 3 20 21 14 22 0
Some Problems: 4 42 36 45 31 55 0
->
J 8 1 1 8 20 1 1 0
2
5 2 J 1 1 0
Major Problems: 1 0 0 0 9 0 0
In addition to asking respondents to give a general rating of their memory ability,
both surveys ask subjects to compare their memory ability to how it was one year ago.
Eightv-five percent of the AHEAD sample indicated that their memory ability was the
same as it was last year. 12% said their memory ability was worse, and 3% said their
memory ability was better. The same pattern is observed in the LBLS sample where
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
73
65% of respondents rated their memory ability as the same, and an additional 24% of
respondents selected the responses surrounding the response for a total o f 86% o f the
respondents. Three percent o f the respondents in the LBLS sample said their memory
ability was Much Better, and 3% reported that it was Much Worse.
Table 27. Self-Reported Memory Ability 1 Year Comparison by Sex in the AHEAD.
AHEAD n = 6436
Men Women Total
Better: 3 2% 3% 3%
Sam e: 2 85 85 85
Worse: 1 13 12 12
Table 28. Self-Reported Memory Ability I Year Comparison by Sex in the LBLS.
LBLS n = 230
Men Women Total
Much Better: 7 1% 5% 3%
6
2 2 2
5 13 10 12
Same: 4 66 63 64
. > 11 10 10
2
7 6 6
Much Worse: 1 1 5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
74
It was possible to compare how subjects in both samples scored on the free recall
performance tasks by looking at the proportion correct. As seen in Table 29. the mean
proportion correct was .51 o f the 20 word visual list in the LBLS study, and .45 o f the 10
word auditory list in the AHEAD study. In both cases, the standard deviation was .19.
Women outperformed men. in both samples. The mean score for women in the AHEAD
sample was .47 which was .04 higher than the men. In the LBLS sample the mean
proportion correct for women was .55 which was .06 higher than the men.
Table 29. Mean Proportion Correct in List Recall by Sex in the AHEAD and LBLS Data.
AHEAD n=6436 LBLS n=230
Mean SD Mean SD
Men .43 .18 .49 .17
Women .47 .19 .55 .19
Total .45 .19 .51 .19
In both samples, the mean proportion of words correctly recalled diminished with
increasing ages groups (Table 30). In the AHEAD sample the youngest age group, ages
70-74. remembered .51 or approximately 5 words, while the oldest group ages 95 and
older, recalled .25 or between 2 and 3 words. Between each of the 5 year age groupings,
the difference in memory performance was approximately .05. In the LBLS sample, this
trend is apparent, though less pronounced. The youngest age group, ages 70-74.
remembered .58 o f the words or nearly 12 words, while the second oldest group (ages 90-
94) recalled .40 o f the word list, or 8 words. (The 90-94 group was used for comparison
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
75
because there is only one subject in the 95+ group in the AHEAD sample). Between each
of the 4 age groupings, the difference in memory performance was not as consistent as it
was in the AHEAD sample, ranging from .01 to .10.
Table 30. Mean Proportion Correct in List Recall by Age in the AHEAD and LBLS Data.
AHEAD n=6436 LBLS n=230
Age Mean SD Mean SD
70-74 .51 .18 .58 .17
75-79 .46 .18 .55 .22
80-84 .40 .17 .46 .17
85-89 .36 .17 .47 .15
90-94 .31 .19 .40 .13
95+ .25 .13 .50 n/a
In order to determine whether self-reported memory ability predicts objective
memory performance in either of the data sets, hierarchical regression analyses of the
memory performance score were conducted with predictor variables entered in blocks.
In Model 1. the demographic characteristics were entered in the first block, health
indicators in the second, depression in the third, and self-reported memory ability in the
fourth. In Model 2. the two measures of self-reported memory ability were entered in the
first block, depression in the second, self-rated indicators of health in the third, and the
demographic characteristics were entered last. Within each o f the predictor blocks the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
76
variables were variously ordered to examine their explanatory contribution and possible
interaction with other independent variables.
Table 31 summarizes the results o f the first hierarchical regression equation. The
unstandardized regression coefficients (5) are presented in the first column. These
coefficients are estimates o f the change in the dependent variable that can be attributed to
a change of one unit in the independent variable. The unstandardized regression
coefficients reflect the results o f the model when all the variables have been entered. The
second column presents the additional amount o f variance explained (R- Chg) when each
variable is entered. In the AHEAD sample, every predictor variable was significant and
the overall equation explained 24% of the variance in performance on the list recall task.
Age accounted for 9% o f the variance in memory ability. The unstandardized regression
coefficient indicated that the association was in the expected direction, with older age
predicting lower performance on the memory task. After partialling out age. sex
accounted for an additional 1% of the variance, indicating that females recall better than
males. Education accounted for 11% of the variance in memory ability with each year of
schooling associated with a .0155 increase in the proportion correct. In this model, the
demographic characteristics accounted for a very substantial, 22% of the variance. After
partialling out the demographic characteristics, the self-reported indicators of health were
entered, starting with self-rated health, then self-rated eyesight, and then self-rated
hearing. Each accounted for less than 1% o f the variance. Reporting better health,
eyesight, and hearing was associated with a better performance score on the list recall
task. Taken together the self-rated indicators o f health accounted for an additional 2% of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
77
the variance. Depression was entered in the third block and explained less than 1% o f the
variance in objective memory ability. Being more depressed was associated with worse
performance on the memory task. Having partialled out sex, age, education, health status,
and depression, the final block introduced the independent variable o f interest, general
self-reported memory ability. General self-reported memory ability was significant in the
AHEAD data. More positive assessment of memory ability was associated with better
performance on the list recall task, but the self-reported memory explained less then 1%
o f the variance in memory performance.
When the same model was run on the LBLS convenience sample, it accounted for
16.5% o f the variance. The only two variables that were significant predictors of
objective memory' performance, were age and sex. Age accounted for approximately 9%
o f the variance, with older age predicting worse performance on the memory task. Sex
explained 4% of the variance, and like the AHEAD sample, being a female was
associated with better memory performance. Education, which was such an important
predictor in the AHEAD sample, was not significant in the LBLS sample, nor was
general self-reported memory ability.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
78
Table 3 1. Results of the Model 1 Hierarchical Regression Analysis o f the Predictors of
Objective Memory Ability in the AHEAD and LBLS Data Sets
AHEAD n=6461 LBLS n=230
h R-'Chg b R=Chg
Intercept .7683 — .9710 —
Age -.0084** .0954** -.0095** .0875**
Sex .0451** .0121** .0681* .0395*
African American -.0658** .0348** -.1215 .0022
Hispanic -.0019 .0088** -.0768 .0062
Southern Region -.0135* .0051** — —
Education .0140** .1751** .0050 .0082
Self-Reported
Health
.0052* 0074** .0154 .0034
Self-Reported
Eyesight
.0050* .0027** -.0142 .0051
Self-Reported
Hearing
.0107* .0044** .0057 .0011
Depression -.0058** .0037** .0035 .0058
Self-rated
Memory
.0064* .0016** .0144 .0109
Overall Rsq .2523 .1690
In order to examine how the order entry o f the predictor blocks impacts the
outcome, the order of entry of the blocks, and the variables within the blocks, were varied
in Model 2. Table 32 presents the outcome of the Model 2 hierarchical regression
analyses. Only the R- Changes are presented as the unstandardized regression
coefficients remain the same. Again, all the of the variables were significant predictors in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
79
the AHEAD sample. General self-reported memory ability accounted for 2% o f the
variance. Next, depression was entered which accounted for an additional 2% o f the
variance. Self-rated hearing, eyesight and health, were entered next and accounted for
2%. 1%. and less than 1%, respectively. When entered in the last block, the demographic
characteristics did not account for nearly as much variance as they accounted for when
they were entered in the first block. Sex explained less than 1%. as did education (which
explained 11% in Model I) and age accounted for an additional 5.5% after the effects of
all the other variables were partialled out. As opposed to Model 1. where the
demographic characteristics explained 21% o f the variance, the demographic
characteristics accounted for only 7% o f the change in the R2 . This suggests that that
self-reported health indicators share variance with the demographic characteristics. In the
LBLS sample, self-reported memory ability was significant in Model 2. explaining 2% of
the variance. In the LBLS sample, the demographic characteristics sex. age and
education were all still significant, together accounting for 12% of the variance.
Specifically, sex accounted for 2%, education accounted for 2% and age accounted for
8% of the variance.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
80
Table 32. Results of the Model 2 Hierarchical Regression Analysis of the Predictors of
Objective Memory Ability in the AHEAD and LBLS Data Sets
AHEAD LBLS
R2 Change R2 Change
Self-Reported
Memory Ability
.0220** .0197*
Depression .0211** .0039
Self-Reported Hearing .0184** .0083
Self-Reported Eyesight .0102** .0022
Self-Reported Health .0044** .0027
Sex .0070** .0226**
Education .00872** .0176*
Age .0558** .0800**
Southern Region .0017** --
Hispanic .0001 .0047
African American .0101** .0017
Overall R2 .2392 .1690
While the finding that self-reported memory ability was significant predictor of
memory ability in the AHEAD sample but not in the LBLS sample may reflect the
importance of sample composition, it could also reflect the statistical power associated
with large samples. In order to examine this possibility, an analyses from a computer
generated random sample of 230 subjects from the AHEAD sample was conducted in
order to match the number o f subjects in the AHEAD and LBLS samples. The AHEAD
subsample was very similar in composition to the AHEAD full sample (Table 33).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
81
Table 33. Descriptives o f the Full AHEAD Sample and a Subsample o f the AHEAD.
AHEAD
FULL SAMPLE
n=6451 AHEAD
SUBSAMPLE
n=230
Mean SD Mean SD
Age 77.06 5.53 77.15 5.93
Sex 63 % Female n/a 66% Female n/a
African American 10% n/a 7% n/a
Hispanic 4% n/a 4% n/a
Southern Region 32% n/a 35% n/a
Education 11.18 3.50 11.50 n/
Self-report Health 3.03 1.15 2.99 1.14
Self-report Vision 3.20 1.09 3.21 1.11
Self-report Hearing 3.27 1.07 3.25 1.09
Depression 1.61 1.95 1.77 1.88
Self-Rated Memory 3.16 .99 3.13 .99
Memory Ability .46 .19 .47 .17
Having established the comparability of the AHEAD full and subsamples, the
Model 1 hierarchical regression analysis was conducted on the AHEAD subsample. Table
34 compares the results o f the analysis run on the subsample to the results of the same
analysis run on the full AHEAD sample and the full LBLS sample. The results o f the
hierarchical regression model using the AHEAD subsample are more similar to the LBLS
results than the full sample AHEAD results. The R squared change and beta coefficients
indicate that self-rating o f memory does not reflect memory perforamnce.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
82
Table 34. Results o f the Model 1 Hierarchical Regression Analysis of the Predictors o f
Objective Memory Ability in the full AHEAD Sample, LBLS sample, and AHEAD
subsample.
AHEAD n=6461 LBLS n=230 AHEAD n=230
h R: Chg b R- Chg b R- Chg
Intercept .7683 — .9710 — .4784 —
Age -.0084** .0954** -.0095** .0875** -.0054* .0715**
Sex .0451** .0121** .0681* .0395** -.0602* .0249*
African
American
-.0658** .0348** -.1215 .0022 -.0516 .0270*
Hispanic -.0019 .0088** -.0768 .0062 -.0093 .0056
Southern
Region
-.0135* .0051** — — .0105 .0001
Education .0140** .1751** .0050 .0082 .0140* .0745**
Self-
Reported
Health
.0052* .0074** .0154 .0034 -.0051 .0011
Self-
Reported
Eyesight
.0050* .0027** -.0142 .0051 .0178 .0121
Self-
Reported
Hearing
.0107* 0044** .0057 .0011 .0018 .0004
Depression -.0058** .0037** .0035 .0046 .0031 .0006
Self-rated
Memory
.0064* .0016** .0144 .0109 .0185 .0108
Overall R- .2523 .1690 .2295
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
83
C. THE INCLUSION OF PROXY RESPONDENTS
Eleven percent of the AHEAD sample (n=768) had proxy respondents and a
greater proportion of men had proxy respondents (13%) than did (9%) women. Proxy
respondents increased with age, ranging from 8% o f subjects between the ages o f 70-79
to 55% for those age 95+ (Table 35).
Table 35. Use o f Proxy Respondent by Age Groups in the AHEAD Data Set.
AHEAD n=7219
70-74 75-79 80-84 85-89 90-94 95+
Use Proxy 8% 8% 12% 15% 29% 55%
Table 36 compares the age, sex, education, health, and sensory status o f the proxy
respondents as compared to the self-respondents. The mean age of subjects with proxy
respondents was 80 years old which is three years older than the mean age of the self
respondents. There are approximately equal numbers of men and women in the proxy
sample as compared to the self-respondents, where women account for nearly 70% of the
sample. The mean education o f the proxy respondents is 9 years as opposed to 11 years
in the self-repondent sample. The rating o f health and sensory ability was lower for
those with proxy respondents than for those who rated themselves.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
84
Table 36. Means and Standard Deviations of the Characteristics of Proxy Respondents
and Self-Respondents.
AHEAD Proxy Respondents ,
(n=768)
Self-Respondents
(n=6436)
Age 80.10(7.28) 77.06 (5.53)
Sex 48% Male 36% Male
African American 14% 10%
Hispanic 8% 4%
Northern Region 21% 16%
Southern Region 23% 32%
Eastern Region 37% 40%
Western Region 18% 16%
Education 9.04 (4.04) 11.18(3.50)
Health Status 2.37(1.21) 3.03 (1.15)
Eyesight 2.64(1.14) 3.20(1.09)
Hearing 2.71 (1.14) 3.27(1.07)
AHEAD interviewers were instructed to attempt to interview all selected
respondents but to accept the respondent's or a caretaker's judgement that the respondent
is unable to participate in the interview. Interviewers had the option to terminate the
interview if it became clear that the data would be o f low quality. In making this
determination, interviewers were told to look for: inconsistent answers; many "Don't
Know" responses: very slow progression on the interview; and poor performance on the
cognitive tests. Such termination was usually not necessary because in most cases, proxy
interviews were conducted from the beginning (Herzog, 1996).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
85
In order to learn more about the proxy selection process, Herzog (1996) reviewed
the field notes of the interviewers. According to these notes, 72% of proxy respondents
were chosen because the selected respondent was unable to participate in the interview
because o f health problems, sensory impairment, and/or cognitive impairment. Twenty-
two percent o f the proxies responded tor a selected participant who was unwilling to
participate, but willing to let someone else answer the questions for him. A few proxy
respondents (6%) were chosen because the respondent was not sufficiently proficient in
English or Spanish. Table 37 summarizes the relation of the proxies to the participants
they responded for. As expected, the largest proportions o f proxy respondents were
spouses (44%). then daughters (28%), then sons (12%).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 37. Relation o f Proxy Respondents to the Selected Participant
86
AHEAD n=768
Spouse 337 44%
Daughter 217 28%
Son 89 12%
Daughter-in-law 25 3%
Son-in-law 8 1%
Granddaughter 17 2%
Grandson 7 1%
Sister 14 1%
Brother 1 2%
Other Relative 21 <1%
Friend 18 3%
Paid Helper 9 1%
Professional 5 1%
Next, it was possible to examine how the proxy respondents rated memory ability.
Among proxy respondents, the mean rating o f memory ability was 2.58 with a standard
deviation of 1.28. as compared to self-respondents who had a mean memory rating o f
3.16 with a standard deviation of .99. Table 38 compares the responses o f the proxy
respondents to the responses of the self-respondents. In both groups. 9% o f the
respondents selected the highest rating. In contrast, 26% of the proxy sample rated the
memory o f the proxy as Poor as compared to only 4% of the self-respondents. While the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
87
difference in rating may reflect differences in proxy reports as opposed to self-reports, it
also undoubtedly reflects the fact that those who use proxy respondents are more likely to
be in worse health, and more cognitively impaired.
Table 38. Self and Proxy Ratings of Memory Ability for Each Response Option.
AHEAD n=72l9
Proxy Respondents Sel f-Respondents
Excellent: 5 9% 9%
Very Good: 4 15% 26%
Good: 3 16% 40%
Fair: 2 23% 20%
Poor: 1 26% 4%
To examine the possibility that proxy ratings may vary by relationship o f the
proxy respondent, the proxy relationships were collapsed into 6 categories: spouse, child
(son and daughter), in-law (son-in -law and daughter-in-law), friend, paid helper, and
other. Among spouse proxy respondents, the mean memory rating was 2.58. among adult
children it was 2.60. for in-laws it was 2.49, for friends it was 1.98, for helpers it was
2.10. and the mean for the "other” category was 2.68. Table 39 summarizes the proxy
ratings o f memory ability by relationship o f the proxy respondent with the selected
participant. T-tests indicated that the difference between the ratings o f adult children and
spouse was not statistically significant. The only group that was statistically different
from the rest was the "friends” group who rated memory ability more negatively.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 39. Proxy Rating o f Memory Ability by Proxy Relationship
88
AHEAD n=768
Spouse Child In-Law Friend Helper Other
Excellent 6% 13% 9% 6% 7% 8%
Very Good 17 14 21 6 0 15
Good 30 22 18 6 21 39
Fair 23 25 15
*> - t
21 11
Poor 24 25 38 44 43 28
D. MEASUREMENT OF SELF-REPORTED MEMORY ABILITY
In addition to the general rating of memory ability and rating o f memory ability
compared to one year ago, the MFQ queries respondents on 62 other items that comprise
4 unit-weight factor scales: Frequency o f Forgetting, Seriousness o f Forgetting,
Retrospective Functioning, and Mnemonics Usage.
The item that is most similar to the single item self-rating o f memory is the
factor. Frequency o f Forgetting. In the LBLS sample, the mean rating on the general
measure of self-rated memory ability was 4.61 and the mean rating Frequency of
Forgetting factor was 4.74, in both cases the rating is closest to Sometimes. The
correlation between the single item general rating o f memory and the Frequency of
Forgetting score is statistically significant at .56. The item that is closest to the single
item rating of memory ability 1 year ago is the factor. Retrospective Functioning. Both
the single item comparison and the factor Retrospective Functioning had mean scores of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
89
3.93. The correlation between the rating o f memory ability compared to one year ago and
Retrospective Functioning was also significant at .37. There was no equivalent
abbreviated measure for the factors Seriousness o f Forgetting or Mnemonics Usage. The
mean o f score on Seriousness of Forgetting was 4.72 with a standard deviation o f 1.22. a
rating indicating problems are considered to be Somewhat Serious. The mean o f
Mnemonics Usage was 3.13 with a standard deviation o f 1.4, suggesting that mnemonics
techniques are a little more than Sometimes.
Using the equations previously run on data from both data sets, regression
analysis was conducted to determine whether any additional variance was explained when
the LBLS equation was run with more extensive measures o f self-reported memory
ability. The analyses were only conducted on the LBLS sample because the AHEAD
data set did not include the more comprehensive measure o f self-reported memory ability.
The MFQ factor. Mnemonics Usage, was not included in the final analyses because it did
not account for any additional variance, so using the three factors only resulted in a more
parsimonious model. As with the analyses o f the abbreviated measures, two models were
used to examine the relative contribution o f the predictor variables to the objective
memory performance. Table 40 summarizes the results o f Model 1. The first two
columns present the results of the hierarchical regression analyses of the abbreviated
measures. The next two columns summarize the analyses of the full MFQ factors.
Frequency o f Forgetting. Retrospective Functioning, and Seriousness of Forgetting.
Immediately apparent is the fact that the overall R2 for the equation using the MFQ
measures is less than 1 percentage point higher than the R2 of the abbreviated measure.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
90
Yet. unlike the analyses o f the abbreviated measures, in the analyses of the full measures,
self-rated Frequency o f Forgetting was a significant predictor o f memory ability even
after controlling for age, sex, education, self-rated health indicators, and depression. The
analysis using the MFQ indicates that self-reported memory predicts memory ability
performance, but it only accounts for three percent o f the variance in memory ability. In
contrast, age accounts for 9% o f the variance and sex accounts for 4%. The three
significant independent variables, age, sex, and self-reported Frequency of Forgetting, are
related to self-reported memory ability in the expected direction. Being older, male, and
indicating that you have more frequent memory problems, predicts worse performance
on the list recall task in this sample.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
91
Table 40. Results o f the Model 1 Hierarchical Regression Analysis o f the Predictors o f
Memory Performance on a List Recall Task Using Abbreviated and Full Measures o f
Self-Reported Memory Ability.
LBLS
N=230
Abbreviated
Memory
Measure
Abbreviated
Memory
Measure
b R' Chg b
R2 c h Z |
Intercept .9710 —
1.025
Age -.0095** .0875** -.0098** .0932**
Sex .0681* .0395* .0673* .0362*
African American -.1215 .0022 -.0990 .0023
Hispanic .0768 .0062 -.0670 .0064
Education -.0050 .0082 .0038 .0078
Self-Reported Health .0154 .0034 .0202 .0025
Self-Reported Eyesight -.0142 .0051 -.0124 .0046
Self-Reported Hearing .0057 .0011 .0050 .0007
Depression .0035 .0046 .0037 .0052
Self-rated Memory -.0144 .0109 — —
Frequency o f Forgetting — -- .0287* .0072
Retrospective
Functioning
— — .0097 .0042
Seriousness of
Forgetting
-- -- .0139 .0052
Overall R : .1690 .1754
To examine how the order o f entry o f the predictor blocks impacts the outcomes,
the order o f entry of the blocks, and the variables within the blocks, were varied in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
92
Model 2. In Table 41 only the R2 changes are presented because the regression
coefficients remain the same. The order o f the entry of the variables impacted the
analyses of the abbreviated measure o f self-reported measure which was significant when
entered before sex and education. In the Model 2 analyses of the full measures o f self-
reported memory ability, the results were almost identical despite the variation in the
order o f entry, suggesting that they share less variance.
Table 41. Results of the Model 2 Hierarchical Regression Analysis o f the Predictors of
Objective Memory Using Abbreviated and Full Measure Rating o f Self-reported
Memory.
LBLS
n=230
Abbreviated
Memory
Measure
Abbreviated
Memory
Measure
R2 Change R2 Change
Seriousness o f Forgetting — .0084
Retrospective Functioning — .0082
Frequency o f Forgetting — .0149*
General Self-Reported Memory .0197* —
Depression .0039 .0021
Self-Reported Hearing .0083 .0094
Self-Reported Eyesight .0022 .0005
Self-Reported Health .0027 .0074
Sex .0226** .0248*
Education .0176* .0116
Age .0800** .0833**
Hispanic .0047 .0035
African American .0017 .0012
Overall R- .1690 .1754
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
93
E. MEASUREMENT OF OBJECTIVE MEMORY ABILITY
Just as the adequacy o f the measurement o f self-reported memory may affect the
appearance o f its relationship with objective memory performance, the adequacy o f the
measure o f objective memory performance may also impact the appearance o f the
relationship. In addition to the free recall task, the LBLS study includes two measures of
prose recall. In the prose recall tasks, subjects read an approximately 230 word essay,
while listening to it aloud. Participants are instructed to write down anything they can
remember from the essay. Interviewers encourage participants to recall the text verbatim
if possible, or else to include anything that they remember in their own words. Prose
recall is scored by parsing die passage into content units and relations between the
contents units. The score consists of the proportion of correctly recalled content and
relation units. Table 39 presents the mean proportion correct in the list recall task, the
first prose recall task (Prose I), the second prose recall task (Prose 2). and the combined
measure which is the mean score of all three tasks. On the list recall task, participants
recalled approximately 10 words (51%) from the 20 word list, with a standard deviation
o f . 19. There were subjects who could not remember any of the words, as well as
subjects who recalled all 20. As summarized in Table 42, on Prose Recall 1. the mean
score was .27 with a standard deviation of .12. A score o f .27 reflects that subject
correctly recalled 27% o f the possible 97 content and relational units, or 26 units. The
highest scoring participant on Prose recall 1 recalled 61% of the content units o f the
essay, and the lowest scoring participant did not recall any of the essay. The mean score
on Prose Recall 2 was .30 with a standard deviation o f .16. Again the lowest score was
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
94
0. but the highest score was somewhat higher at 80% correct. The mean score o f the
combined measure was .36 with a standard deviation o f .13. The lowest score was .06
and the highest was .76.
Table 42. A Comparison of the Measures of Memory Performance in the LBLS Data Set.
LBLS n=230
Proportion
Correct
SD Minimum
Score
Maximum
Score
List Recall .51 .19 .00 1.00
Prose Recall 1 .27 .12 .00 .64
Prose Recall 2 .30 .16 .00 .80
Combined Measure .36 .13 .06 .76
The aim of the next set of analyses was to examine how the results of the
regression analyses change when the abbreviated measures o f self-reported memory
ability are used to predict the more extensive measure o f objective memory performance.
For the purpose of comparison, the first two columns o f Table 43 review the results of the
hierarchical regression analysis on the abbreviated measure o f memory ability (list
recall), and the abbreviated measure of self-reported memory ability (general self report)
were run in earlier analyses. In the last two columns, the results of the hierarchical
regression analysis on the more extensive memory measure (list and prose recall) and the
abbreviated measures of self-reported memory ability are presented. As hypothesized,
the overall R- increased when the more extensive measure o f objective memory was used.
The overall R2 in the model predicting the abbreviated measure of memory explained
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
95
16.5% o f the variance while the model predicting the more extensive measure explained a
little over 24% o f the variance, for an increase of 7.5%. In the model predicting the
abbreviated memory measure, self-reported memory ability was not a significant
predictor. In contrast, self-reported memory ability was significant when used to predict
the more extensive measure o f memory ability. After partialling out the effects of age.
sex. education . the self-reported indicators of health, and depression, the single items of
self-reported memory ability explained a little less than 2% o f the variance in
performance on the combined list and prose recall score. Rating your memory more
positively was associated with higher performance on the memory tasks.
Table 43. Results of the Hierarchical Regression Analysis o f the Predictors of the
Abbreviated and More Extensive Measure o f Objective Memory Ability.
LBLS
N=230
Abbreviated
Memory
Measure
Extensive
Memory
Measure
b R2 Chg b R'-Chg
Intercept .9710 -- .8308 —
Age -.0095** .0875** -.0083** .1609*
Sex .0681* .0395* .0326* .0129*
African American -.1215 .0022 -.0980 .0052
Hispanic .0768 .0062 -.0702 .0111
Education -.0050 .0082 .0060 .0257*
Self-Reported Health .0154 .0034 .0073 .0067
Self-Reported Eyesight -.0142 .0051 -.0027 .0010
Self-Reported Hearing .0057 .0011 -.0044 .0005
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
96
LBLS
N=230
Abbreviated
Memory
Measure
Extensive
Memory
Measure
Depression .0035 .0046 -.0003 .0001
Self-rated Memory -.0144 .0109 .0150* .0196*
Overall R- .1690 .2439
Table 44 presents the results o f the analysis when the more extensive MFQ
measures of self-reported memory ability are used to predict the more extensive measure
of objective performance. Again, the first two columns are included for comparison as
they present the results of the hierarchical regression analysis using the more extensive
measure of memory ability and the abbreviated measures of self-reported memory ability
from the preceding analysis. The last two columns present the results o f the hierarchical
regression analysis on the more extensive memory measure (list and prose) and the more
extensive measures o f self-reported memory ability (MFQ factors). Contrary to
expectation, the overall R2 was not increased when the more extensive measure of
objective memory was used, in fact, the analyses using the more extensive measure of
self-reported memory ability explained slightly less of the variance in list and prose
recall.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
97
Table 44. Results o f the Hierarchical Regression of the Predictors o f the More Extensive
Measure o f Objective Memory Using the Extensive and Abbreviated Measures o f Self-
Rated Memory.
LBLS
N=230
Extensive
Memory
Measure
Extensive
Memory
Measure
b R ' Chg b R2 Chg
Intercept .8308 — .8449 —
Age -.0083** .1609* -.0086** .1663*
Sex .0326* .0129* .0320* .0119*
African American -.0980 .0052 -.1063 .0053
Hispanic -.0702 .0111 -.0660 .0113
Education .0060 .0257* .0055* .0263*
Self-Reported Health .0073 .0067 .0130 .0072
Self-Reported Eyesight -.0027 .0010 -.0028 .0009
Self-Reported Hearing -.0044 .0005 -.0034 .0003
Depression -.0003 .0001 -.0003 .0001
Self-rated Memory .0150* .0196* — —
Frequency o f Forgetting — — .0202* .0105
Retrospect. Functioning — — -.0064 .0040
Seriousness o f Forgetting — — -.0066 .0025
Overall R- .2439 .2464
When the order of entry was varied, age, education, and self-reported memory
remained significant (Table 45). Both the general rating and the Frequency of Forgetting
factor explained an additonal 2% o f the variance in memory ability when entered before
age. bringing them up to nearly 4% o f the variance.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
98
Table 45. Results o f the Model 2 Hierarchical Regression Analysis o f the Predictors o f
Objective Memory Using Abbreviated and Extensive Measures o f Self-Reported
Memory Ability.
LBLS
n=230
Extensive
Memory
Measure
Extensive
Memory
Measure
R2 Change R2 Change
Seriousness of Forgetting — .0004
Retrospective Functioning -- .0061
Frequency of Forgetting -- .0367*
Self-Reported Memory .0413* —
Depression .0009 .0025
Self-Reported Hearing .0008 .0026
Self-Reported Eyesight .0001 .0010
Self-Reported Health .0004 .0063
Sex .0069 .0090
Education .0443* .0350*
Age .1320* .0136*
Hispanic .0084 .0073
African American .0024 .0030
Overall /?- .2439 .2464
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
99
CH APTER V: ACCURATE VERSES INACCURATE
SELF-APPRAISAL O F MEM ORY ABILITY
The results o f the Hypothesis One models indicated that self-ratings o f memory
performance are not strongly associated with objective performance on a list recall task.
This finding leads to a related question: are there systematic differences in the subjects
who accurately assess their memory ability and the subjects who make inaccurate
appraisals? Specifically, are there systematic differences between those who over
estimate their memory ability, those who under-estimate their memory ability, and those
who accurately appraise their memory ability? Table 46 offers a heuristic depiction of
this question, in which the "+ ” represents accurate estimation. "UE ‘"represents under
estimation. and “OE" reflects over-estimation of memory ability.
Table 46. A Model o f the Accuracy o f Self-Ratings o f Memory Ability.
Good
Memory Ability
Average
Memory Ability
Poor
M em ory Ability
Good Self-Rating +
OE OE
Average Self-Rating UE
+
OE
Poor Self-Rating UE UE
A. DESCRIPTIVES
In order to examine the extent to which particpants in the AHEAD and LBLS
samples over- or under-estimated memory ability, it was necessary to recode self-
reported memory ability and objective memory performance into three categories. Table
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
47 presents the recoding strategy for both data sets. The five-point scale for the general
rating o f memory item on the AHEAD was reduced to a three-point scale by combining
the top two ratings into a category labeled “good”. This group represented 35% o f the
sample. The middle rating, representing 40% o f the sample, remained the mid-rating, and
was labeled "average”. The lowest two ratings were combined into the single rating,
"poor" which represented 24% of the sample. In the LBLS data set, the top three
rankings represented 27% o f the sample and were combined into the category labeled
"good". The middle two ratings, or 57% of the sample were comibined into the
"average” category, while the lowest three ranking (16%) were combined into a category
labeled "Poor".
Turning next to the recoding of memory performance, those who recalled 60% of
the word list were grouped as having “good” objective memory performance. In the
AHEAD. 28% o f the sample fell into this category as opposed to 36% in the LBLS
sample. Next, those who recalled 40-50% of the words were recoded as having
"average" memory perforamnce. The “average” group encompassed 44% of the AHEAD
sample, and 41% o f the LBLS sample. Those who could recall 30% or less were grouped
into the "poor" performance category, which represented 28% o f the AHEAD, and 25%
of the LBLS.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 47. Recoding the AHEAD varaibles into 3 Category Variables.
101
AHEAD LBLS
General Self- How would you rate your memory How would you rate your memoiy
Rating of ability? ability?
Memory
Ability
Excellent =5(9%) v 3=Good No Problems =7( 10%)'» 3=Good
Very Good =4(26%) ^ (34%) =6(17%)'* (27%)
Minor Prob = 5(I7% )'» 2=Avg
Good =3(40%) -*• 2=Avg =4(40%) .■ * (57%)
(40%)
Major Prob =3(1 l% N
Fair =2(20%) l=Poor =2(4%) l=Poor
Poor = 1(4%) J> (24%) = 1( 1% )'* ( 16 %)
Objective Please repeat any o f the 10 words Please write down any o f the 20
Memory that you recall. words that you recall.
Ability
Proportion Frequency Proportion Frequency
Correct Correct
1.0 1%V 1.0 l% V
.9 2% 3= Good .9 3% 3=Good
.8 3% (28%) .8 5% (36%)
.7 7% .7 11%
.6 15%-* .6 16%'*
.5 22%'» 2=Average .5 21% ^ 2=Average
.4 22%**' (44%) .4 20% ^ (41%)
.3 15%s. .3 17% ^
2
8% l=Poor
2
6% I =Poor
.1 3% (28%) .1 1% (25%)
0 2%-* 0 \ % *
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
102
Tables 48 and 49. present the cross-tabulations of the recoded three category
variables, self-reported memory and objective memory performance. O f particular
interest are the percentage of subjects that fall upon the diagonal (accurate assessement)
as opposed to those falling in the upper right (over-estimation) and lower left (under
estimation) comers. As summarized in Table 50, over 60% o f the subjects in both
samples were inaccurate in their self-appraisals. Specifically, 27% o f the AHEAD and
31% o f the LBLS under-estimated their ability; while 33% of the AHEAD and 37% of
the LBLS over-estimated their memory ability.
Table 48. Self-rated Memory Ability by Objective Memory Performance in the AHEAD.
Good
Memory
Ability
Average
M emory
Ability
Poor
Memory
Ability
M arginal
Good Self-
Rating
12% 15% 8% 24%
Average Self-
Rating
12% 18% 10% 40%
Poor Self-
Rating
4% 11% 9% 24%
M arginal 29% 44% 28% 100%
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
103
Table 49. Self-rated Memory Ability by Objective Memory Performance in the LBLS.
Good
Memory
Ability
Average
Memory
Ability
Poor
Memory
Ability
Marginal
Good Self-
Rating
12% 10% 12% 27%
Average Self-
Rating
19% 24% 15% 57%
Poor Self-
Rating
5% 7% 4% 16%
Marginal 36% 40% 24% 100%
Table 50. Accurate Assessement. Under-Estimation, and Over-Estimation in the
AHEAD and the LBLS Samples.
AHEAD LBLS
Accurate Assessement 39% 40%
Under-Estimation 27% 31%
Over-Estimation 33% 37%
B. ACCURATE APPRAISAL, UNDER-ESTIMATION, AND
OVER-ESTIMATION
To determine whether there were systematic differences between the participants
who accurately assessed their memory ability and the participants who did not. a
multinomial logistic regression analysis was conducted on the AHEAD data set. The
small size o f the LBLS sample precluded the use of the LBLS data. As discussed in
Chapter III. multinomial logistic regression is able to model the relationship between a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
104
three category dependent variable and categorical and continuous independent variables.
The result o f the equation is a model o f the probability o f under-estimation or over
estimation relative to the probability o f accurate estimation controlling for known
explanatory variables. The results o f the logit model reveal: (1) the extent to which each
independent variables explain the relevant category o f the dependent variable (parameter
estimates): (2) how much impact each independent variable has on the relevent dependent
variable (marginal probability); and (3) an estimate o f the relative effect o f the
independent variable in comparison to the omitted values (odds ratio).
As summarized in Table 51, results o f the multinomial logistic regression
indicate that increasing age is negatively associated with over-estimating memory ability,
and positively associated with under-estimating memory ability. The probability o f a 75
year-old person under-estimating memory ability is 6 percentage points less than that of a
70 year-old. while the probability o f over-estimating memory abilty is 7 percentage
points greater than a 70 year-old. The probability o f an 80 year-old under-estimating
memory ability is 11 percentage points greater than a 70 year-old. and the probability is
for over-estimating is 11 percentage points less. Following this trend, the probability o f
an 85 year- old under-estimating memory ability is 15 percentage points less than a 70
year-old and for over-estimating. 17 percentage points greater. The relative likelihood
(odds ratio) o f under-estimating memory ability decreases by 4% each year and the
relative likelihood of over-estimating memory ability increases by 10% each year.
Turning next to sex. the results indicate that being a woman rather than a man is
not significantly associated with the probability of under-estimating memory ability but
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
105
significantly decreases the probability o f over-estimating memory ability. The
probability o f a woman over-estimating memory ability is 3 percentage points less than a
man. The relative likelihood o f a woman over-estimating memory ability is 85% that of a
man.
Being Black as opposed to White, decreased the probability of under-estimating
memory ability and increased the probability o f over-estimating memory ability. Even
after controlling for education, region, age, health indicators, and depression, the
probability o f a Black person under-estimating memory ability is nearly 8 percentage
points less than a White person. The relative likelihood of under-estimating memory
ability is 76% that of a White subject. Conversely, the probability of a Black subject
over-estimating memory ability is 10 percentage points greater than that of a White
subject and the relative likelihood is nearly one and a half time greater.
While having more education is not signficantly associated with the probability of
under-estimating memory ability, it does decrease the probability of over-estimating
memory ability. The probability of a person with a junior high-school education over
estimating memory abilty is 5 percentage points greater than that of a person with a high
school education. Conversely, the probability o f a person with a college education over
estimating memory is 5 percentage points less than a person with a high-school
education. The relative likelihood of over-estimating memory ability decreased by 5%
for each year of education.
Turning next to the indicators o f health status, the results suggest that a more
positive self-rating of health status decreases the probability of under-estimating memory
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
106
abiltiy. but is not associated with the probability o f over-estimating memory ability. The
probability that a person who self-rates his health as ‘‘poor” would under-estimate his
memory ability is 5 percentage less than the probability of a person who rates his health
as "good". The relative likelihood o f under-estimation decreases by 12% for each point
higher. Similarly, a more positive self-rating of eyesight decreased the probability of
under-estimating memory ability and increases the probability o f over-estimating
memory ability. The probability of a person who rates his eyesight as "poor" would
under-estimate his memory abilty is 5 percentage points less than someone who reports
his eyesight is “good”. Conversely, the probability that a person who rates his eyesight as
"poor” would over-estimate memory abilty is over 6 percentage points greater than
someone who reports his eyesight is "good”. The relative likelihood of under-estimating
memory ability decreases by 9% for each rating higher on the eyesight rating scale and
the relative likelihood o f over-estimating memory ability increases by 10%.
Like self-rated eyesight, a more positive self-rating of hearing decreases the
probability of under-estimating memory ability and increases the probability of over
estimating memory ability. The probability that someone with a "poor" hearing rating
would under-estimate memory ability is 8 percentage points less than someone with
a"good"rating. 6 points higher for over-estimating. The relative likelihood of under
estimating memory ability decreases by 16% for each rating higher on the hearing self-
rating. and the relative likelihood of over-estimating memory ability increases by 10%
rating higher. Surprisingly, number of depressive symptoms was not signficanltv
associated with under- or over-estimating memory ability relative to accurate appraisal.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
107
Table 51. Multinomial Logistic Regression Predicting the Probability o f Under-
Estimating and Over-Estimating Memory Ability Compared to Accurately Rating
Memory Ability. Beta Coefficients [Marginal Probabilities] (Odds Ratios).
Variables Under-
Estimate
Over-
Estimate
Intercept
1.1091* (-3.032) -1.5665*** (.209)
Age
-.0399*** (-.961) .0367*** (1.104)
Age 75 a
[-.0567] [.0531]
Age 80 a
[-.1087] [.1096]
Age 85 a
[-.1549] [.1677]
Sex
.0547 [.0232] (1.056) -.1581** [-.0272] (.854)
Black
-.2747* [-.0757] (.760) .3353*** [.0963] (1.398)
Southern Region
.0154 [.0022] (1.016) .0625 [.0121] (1.04)
Education
.0439 (1.045) -.0480*** (.953)
Jr. High b
[-.0482] [.0559]
College b
[.0518] [-.0523]
SR Health
. 1H3***
(.880) .0542 (.947)
Poor
Healthc
[-.0546] [.0410]
SR Eyesight
-.0847** (.912)
1029***
(.902)
Poor Eyes c
[-.0508] [.0641]
SR Hearing - 1 4 5 5 * * *
(.843) .1021*** (.093)
Poor Hear c
[-.0776] [-.0641]
Depressive
Symptoms
-.0097 (.990) -.0060 (.064)
Depressed c [-.0009] [-.0030]
u Compared to Age 70; b Compared to High School Education;
. Compared to "Good“;d Compared to Not Depressed.
* p<.05; ** p<.01; ***p<.001.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
108
C. TOWARD AN UNDER-ESTIMATOR AND OVER-ESTIMATOR PROFILE
Based upon the results o f the logit model, it is possible to develop a profile o f the
characteristics of someone more likely to under-estimate memory ability and someone
more likely to over-estimate memory ability. As depicted in Table 52. the under
estimator profile is younger (closer to 70). White, more educated, and reports a more
negative appraisal of health status, eyesight, and hearing. Presenting a very diffferent
profile, an over-estimator is more likely to be older, male, Black, less educated, and
reports a more positive self-rating of eyesight and hearing.
Table 52. Characteristic Profiles Associated with an Increased Probability o f Under-
Estmating and Over-Estimating Memory Ability.
UNDER-ESTIMATOR OVER-ESTIMATOR
Younger Older
Male
White Black
More Education Less Education
Lower Health Self-Rating
Lower Eyesight Self-Rating Higher Eyesight Self-Rating
Lower Hearing Self-Rating Higher Hearing Self-Rating
By manipulating each o f the characteristics in the profiles, it is possible to look at
the relative contribution of each characteristic. Table 53 depicts a series o f vignettes,
ranging from Vignette I to Vignette 7. Vignette 1 parallels the profile for an "under-
estimator”. that is. a 70 year-old. White subject who obtained 16 years o f education and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
109
rated his/her health, eyesight, and hearing as poor. As presented in the last column o f
Table 53, the probability that a person with these characteristics would under-estimate
memory ability is .65. In the subsequent vignettes, one of the relevant characteristics was
varied in order to examine how the resulting probability changes. For example, in
Vignette 2. the probability that an 85 year-old White subject with 16 years of education
who rated his/her health, eyesight, and hearing as “Poor’ is only .47. a difference o f . 18
compared to Vignette I. In Vignette 3, when the subject is Black rather than White, the
probability o f under-estimating is .56. Similarly, when the subject has a junior high-
school education rather than a high-school education the probability is .54. For health
status, a "Good" rather than “Poor” reduces the probability to .59. and for eysight the
probability is .60. and for hearing the probability is .57. From these hypothetical
vignettes, we see that age is the most influential characteristic, followed by lower
education, race, self-rated hearing, self-rated health, and self-rated eyesight.
Table 53. The Probability of Under-Estimating Memory Ability Under Seven
Hypothetical Vignettes Each Manipulated to Highlight a Single Characterstic.
Vignette
N um ber
Age Race Educ Self-rated
H ealth
Self-rated
Eyesight
Self-rated
H earing
Prob.
1 70 White 16 yrs Poor Poor Poor .65
2
85 White 16 yrs Poor Poor Poor .47
■ * »
J 70 Black 16 yrs Poor Poor Poor .56
4 70 White 8 yrs Poor Poor Poor .54
5 70 White 16 yrs Good Poor Poor .59
6 70 White 16 yrs Poor Good Poor .59
7 70 White 16 yrs Poor Poor Good .57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 1 0
A comparable comparison was conducted by manipulating the characteristics
asociated with over-estimating memory ability (Table 54). Vignette I parallels the profile
for an '‘over-estimator’', and presents the probability of a 85 year-old. Black male with 8
years o f education who rated his health, eyesight, and hearing as "Good" over-estimating
memory ability. The probability that a person with these characteristics would over
estimate memory ability is .57. In Vignette 2, the subject is a 70 year old rather than 85
years old which lowers the probability o f over-estimating to .39. If the subject was an 85
year-old Black female with 8 years of education and positive health, eyesight, and
hearing rating, the probability would be .53. If the subject is White rather than Black, the
probability of over-estimating is .47. Similarly, when the subject has a high-school
education rather than a junior high education, the probability is .45. A eyesight self-rating
of "poor" rather than "good" reduces the probability to .51, and for hearing, to .50.
Table 54. The Probability o f Over-Estimating Memory Ability Under Seven Hypothetical
Vignettes Each Manipulated to Highlight a Single Characteristics.
Vignette
Num ber
Age Race Sex Education Self-rated
Eyesight
Self-rated
H earing
Prob.
1 85 Black 8 yrs Good Good Good .57
2
70 Black 8 yrs Good Good Good .39
85 W hite 8 yrs Good Good Good .47
4 85 Black 16 yrs Good Good Good .45
5 85 Black 8 yrs Poor Good Good .53
6 85 Black 8 yrs Good Poor Good .51
7 85 Black 8 yrs Good Good Poor .50
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Based upon the preceding results, it is possible to rank the characteristics
associated with over- and under-estimating memory ability. As summarized in Table 55,
age is the most important variable for predicting both types o f memory inaccuracies. Next
in order of importance were education, followed by race, self-rated hearing and self-rated
health. Self-rated health occupies the final positions in under-estimating memory ability,
while sex occupies the final positions for over-estimating memory ability.
Table 55. A Profile of the Characteristics of an Under-Estimator and Over-Estimator
Ranked by Order of Importance.
UNDER-ESTIM ATOR OVER-ESTIM ATOR
Younger - < — ► Older
More Education Less Education
White Black
Low Hearing Rating < — > • High Hearing Rating
Low Eyesight Rating < — > High Eyesight Rating
Low Health Rating Male
With the exception of self-rated health and sex (both o f which held the lowest
ranking), all o f the variables associated with a greater probability of under-estimating
memory ability were inversely associated with a greater probability o f over-estimating
memory ability. Further, the order o f relative importance of these characteristics was
identical for both types of inaccurate appraisals. Those who are younger, that is closer to
age 70. are more likely to under-estimate their memory ability while those who are older
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
112
are more likely to over-estimate memory ability. Having more education is associated
with under-estimating memory ability, while having less education is associated with
over-estimating memory ability. Surprisingly, even after controlling for education, being
White is associated with over-estimating memory ability and being Black increased the
probability of over-estimating memory ability. More negative self-assessments were
associated with under-estimating memory ability and more positive self-assessment
associated with over-estimating memory ability. Self-rated health status followed the
trend with under-estimation but not for over-estimation. Being a male increased the
probability associated with over-estimating memory, but being a female did not increase
the probability of under-estimating memory ability.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
113
CHAPTER VI: DISCUSSION AND IMPLICATIONS
Growing interest in the cognitive functioning of America's senior population has
moved research on memory ability from the exclusive domain of cognitive research into
the multidisciplinary arena of sociological and epidemiogical research. Reflecting this
interest, several nationally representative data sets include cognitive functioning sections
in their survey instruments. Unfortunately however, the time constraints associated with
large, representative samples prohibit comprehensive memory assessment. As a result,
self-reported memory ability has become an increasingly common element o f such
surveys. This aim o f this research was: (1) to determine whether self-reported memory
ability accurately reflects objective memory ability; and (2) to examine whether there are
systematic differences between those who accurately appraise their memory ability as
opposed to those who over-or under-estimate their memory ability.
To avoid the methodological weakness that have compromised previous
investigation in this area, the research design attempted to combine the breadth o f social
epidemiology with the depth o f psychological inquiry by using both a nationally
representative survey sample (AHEAD) and an in-depth cognitive testing sample
(LBLS). Not surprisingly, there were differences in the composition of the two samples.
Reflecting the U.S. population of older adults, most of the AHEAD sample was under the
age of 80. there were many more women than men, 10% of the sample were African-
Americans. and less than a quarter went beyond high school. In contrast, the LBLS
sample had as many participants above age 80 as below, there were more men than
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
114
women. less than 1% were African-American, and more than half o f the sample attended
college or beyond.
The two samples also differed in the way they rated their health status. Whereas
67% o f the AHEAD sample rated their health in the top three ratings, 92% of the LBLS
sample selected the top three ratings. Although the self-ratings of sensory ability did not
differ between the two samples, it must be noted that the AHEAD survey asks
participants to rate their eyesight and hearing with their eye-glasses/hearing aid on.
whereas the LBLS does not. If the participants in the LBLS sample did rate their sensory
ability as it is without the use o f their hearing aids and glasses, then seemingly similar
ratings are in fact different. Similarly, while the results indicate that the AHEAD sample
has higher levels o f depression than the LBLS sample, the conclusion is tempered by the
fact that the operationalizations o f depression are different.
Having compared the composition o f the two samples, it was then possible to
examine how participants in both samples rated their memory abilities. Interestingly . the
pattern of self-ratings was similar between the two samples, and mirrored the memory
self-ratings o f the NHIS sample (n=14,000 ) (Culter & Grams, 1988). In the AHEAD.
LBLS. and NHIS samples, over 75% of the respondents indicated that their memory
ability was Goad/Had Few Problems and nearly 80% indicated that their memory ability
was the Same or Better than it was one year ago. Contrary to Poon's (1985) assertion that,
"the feeling that one's ability to remember and to retrieve information is inadequate is a
universal complaint among middle age and elderly persons" (p.427). the results of all
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
three samples suggest that the majority o f adults over age 70 feel that their memory is
generally good, and the same as it was last year.
With regard to objective performance, the LBLS sample did somewhat better at
remembering the 20 word visual list, than the AHEAD sample did remembering the 10
word auditory list. In both cases women outperformed men on the memory tasks. In
future research, it would be informative to have subjects perform both the 20 word
visual/written free recall task in an in person session, and the 10 word auditory/oral task
over the telephone in order to determine the equivalency of the two tasks. In both
samples, memory performance on the list recall task declined with age. Self-reported
memory also declined with age but the relationship was weaker than the relationship
between age and objective memory performance.
To test the hypothesis that self-rated memory ability does not reflect memory
performance, several hierarchical analyses were conducted. Interest in both the beta and
R squared change reflects the multidisciplinary nature of the subject matter and an
attempt to communicate the results to both the psychological and sociological
community. Interestingly, attention to both components of the analysis provided some of
the most interesting information. Results revealed that while self-reported memory ability
was a signficant predictor of memory ability in the AHEAD data set. it explained only a
small portion o f the variance. In the LBLS sample, neither the beta coefficient nor the R
squared was significant. An analysis was conducted to determine whether the
significance o f the beta coefficient in the AHEAD reflected the importance o f sample
composition, or whether it was primarily a reflection of the statistical power associated
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
116
with large data sets. On the one hand, the convenience sample was older, more educated,
and in better health than the rest o f the population and as indicated in the literature, these
characteristics may impact the relationship between how a person rates her memory and
how well she performs on a memory task. But on the other hand, it may simply reflect the
statistical power associated with large samples in that general self-reported memory
ability explains less than one percent o f the variance in objective memory ability and yet
the beta coefficient is significant. In order to examine whether the differences in the
results of the AHEAD and LBLS sample stemmed from differences in sample
composition or sample size, a computer generated random subsample of AHEAD sample
was selected to be the same size as the LBLS (n=230). Descriptive analyses of the
subsample indicated that its composition was similar to that of the full sample. The
results of the hierarchical regression model run on the subsample were more similar to the
LBLS sample than to the full AHEAD sample. The R squared change and beta
coefficients indicated that neither o f the self-reported memory measures nor self-rated
health, eyesight, hearing, or depression predicted memory ability. Unlike the LBLS
sample, race predicted memory ability even after controlling for education and region.
The findings that being an African American was associated with lower performance on
memory ability is likely to reflect the fact that all of the participants were schooled
between 65- 95 years ago in an era when schooling for African American children was
segregated so that having 8 years o f education for a black respondent was not necessarily
the same amount/quality of education as 8 years of education for a white respondent. In
an attempt to take such differences into account, living in the southern region of the U.S.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
117
was controlled because over 60 years ago, educational differences were likely to have
been more pronounced in the south. The reason that race was not significant in the LBLS
sample was that there were so few African American respondents. With the exception of
race, the findings o f the representative population sample support the findings reported
from a smaller convenience sample.
Using the AHEAD data set. it was possible to consider the possiblity that self-
reports o f memory ability may be an accurate predictor o f memory ability for those who
rated their memory ability as Poor or Fair (24%), or for those who recalled less than 25%
o f the ten word list (mean score was 46% ). Hierarchical regression analyses indicated
that self-reported memory ability was not a significant predictor of objective memory
performance in the subsample o f those with lower self-ratings, nor for those who
perform more poorly on the objective task.
In the next set of analyses, memory ratings o f proxy respondents were examined.
Eleven percent of the AHEAD sample had proxy respondents, with men and older
subjects more likely to have proxies. Those with proxy respondents also tended to be
less educated, in worse health (i.e. the proxy respondents rated the health o f this group
more negatively). Most proxy respondents were spouses, daughters, or sons of the
selected participant and were chosen because the selected respondent was unable to
participate in the interview because o f health problems, sensory impairment, and/or
cognitive impairment. Some were selected because the respondent was unwilling to
participate, but willing to let someone else answer the questions for him. A few proxy
respondents were chosen because the respondent was not sufficiently proficient in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
118
English or Spanish. While the inclusion of proxy respondents is fairly common among
large scale studies, it is rare in cognitive research and thus informative to examine the
characteristics of the participants who would normally not be included in the sample. The
systematic differences between the self-respondents and proxy respondents highlight the
benefits o f incorporating proxy respondents into the research design. While performance
based nature o f cognitive research is not conducive to proxy respondents, it would be
beneficial to use proxy respondents to gain information about the selected participants
who refuse or cannot participate.
In looking at how the proxy respondents rated memory ability, the analyses
indicated that proxy ratings were lower than self-ratings. It is impossible to know to what
extent the difference in rating reflects differences in proxy reports as opposed to self-
reports. because those with proxies are systematically different (more impaired) than
those who respond for themselves. Contrary to expectation, adult children did not rate
memory ability more negatively than spouses. Instead, proxy respondents who were
indentified as "friends" rated memory ability more negatively than the other proxy
relationships. Two explanations are plausible. First, those who use friends as proxy
respondents constitute only 3% of the proxy subsample and may be systematically
different from those who have family members to serve as proxies. If this is the case,
those with friends as proxy respondents may in fact, have worse memory abilities and the
friend ratings reflect this. Alternatively, it may be that those with family proxies are not
systematically different than those who have friend proxies, but instead, that friends are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
119
more objective than family members and are able to give more negative ratings of
memory ability.
Next, the additional measures o f self-reported memory contained in the LBLS
were used to determine whether using a more comprehensive measurement would impact
the outcome. Contrary to expectation, the more extensive measures of the self-reported
memory ability did not account for additional variance in objective memory performance.
The regression coefficient, however, was significant for self-rated Frequency of
Forgetting (and not for general self-rated memory ability). Interestingly. Frequency of
Forgetting was the only factor of the MFQ that was a significant predictor o f list recall.
The analyses could only be run on the LBLS sample as the AHEAD does not contain any
additonal measures of self-reported memory. The next analyses sought to determine how
the results o f the regression analyses changed when a combined list and prose recall task
were used to operationalize the dependent variable, memory ability. The overall R-
increased 7% when the more extensive measure o f objective memory was used. Further,
the regression coefficient for the abbreviated measure of self-reported memory was
significant, whereas it had been insignificant when predicting list recall only. When the
more extensive memory measures were used, the overall R2 did not increase, in fact, the
single item measure of self-rated memory explained more of the variance than Frequency
of Forgetting suggesting that measurement o f memory impacts the outcome more than
the measurement of self-rated memory ability. Taken together, the results o f the
hierarchical analyses of the predictors of memory ability indicate that even after
controlling for sociodemographic characteristics, health indicators, and depression, self
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
120
reported memory ability is only minimally indicative o f objective performance and thus
the null hypothesis can not be rejected.
Turning next to the second research question, whether there are systematic
differences between those who accurately appraise their memory ability as opposed to
those who are inaccurate, it was found that over 60% o f the subjects in both the AHEAD
and LBLS samples were inaccurate in their self-appraisal o f memory ability. O f those,
approximately half under-estimated their memory ability, and half over-estimated their
memory ability. Results of the multinomial logistic regression indicated that there are
systematic differences between those who accurately assess memory ability, those who
under-estimate memory ability, and those who over-estimate memory ability. Being
younger, having more education, being white, and reporting a more negative sensory
status increases the probability of under-estimating memory ability. Conversely, being
older, having less education, being black, reporting more positive sensory and health
status, and being male increases the probability o f over-estimating memory ability. For
both under-estimating and over-estimating age is the most important, followed by
education, race, self-rated hearing and self-rated health. Self-rated health occupies the
final positions in under-estimating memory ability, while sex occupies the final positions
for over-estimating memory ability. Of particular interest is the fact that the independent
variables consistently impact the two types of inaccuracies in the opposite direction.
Table 54 compares the under-over estimator profile to established (and duplicated in the
AHEAD sample) findings on better and worse objective memory performance.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
121
Table 56. A Comparision o f the Characteristics Associated with Under- and Over-
Estimating Memory Ability and the Characteristics Associated with Better and Worse
Memory Performance.
UNDER O V ER BETTER W ORSE
ESTIM ATOR ESTIM A TO R M EM O RY M EM O RY
PERFOM ANCE PERFORM ANCE
Younger
< - * •
Older Younger Older
More Education
4— ►
Less Education More Education
* -* >
Less Education
White
4—*•
Black White
4— ►
Black
- Hearing
4— ►
+ Hearing + Hearing
« -* ■
- Hearing
- Eyesight
« - * •
+ Eyesight + Eyesight
4— ►
- Eyesight
- Health + Health
4- >
- Health
Male Female
4— ►
Male
Age. education, and race produce the same profile in the memory performance
profile as in the under- over-estimator profile. Younger, more educated, white adults tend
to do better on memory performance tasks and are also more likely to under-estimate
memory ability. With better scores, the potential for appraisal error is in under-estimating
memory ability, while for those with low score the potential for appraisal error is over
estimating ability. Expectations and reference groups are likely to contribute to the
inaccurate appraisals. An older adult may over-estimate his ability because he considers
his memory to be better than his peers. Similarly, a highly educated person may under
estimate her ability because she considers anything less than complete retention to be a
memory' problem. Cognitive impairment may also contribute to the preponderance o f
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
122
inaccurate memory appraisal. This phenomenon has been documented in dementia
literaure in that demented patients are more inclined to say that they have no problems
than non-demented patients. Turning next to self-reported health, eyesight, and hearing, a
different pattern emerges. Rather than the two profile parelleling each other, they are the
inverse. This finding is likely to reflect the fact that like memory appraisal these variables
are self-rated. Those who negatively assess their health are more likely to negatively
assess their memory ability even in the absence of memory problems, thus under
estimation. Similarly, those who rate their memory as "good” when it is in fact "poor”
(over-estimate), are also likely to report that their health is "good” even though it may not
be.
The findings presented here suggest that self-reported memory ability is an
inappropriate measure of memory ability because the self-ratings are only minimally
indicative of objective memory performance. If self-ratings o f memory ability do not
predict memory ability, should self-reported memory ability continue to be included in
the cognitive functioning sections o f epidemiological research? The answer depends upon
the goal of the investigators. If the intent is to gauge the memory ability o f older adults,
the designers of epidemiological surveys would be better advised to allot scarce survey
time to objective measures o f memory performance rather than self-ratings o f memory
ability. In the AHEAD survey, memory ability is currently operationalized using an
auditorially administered, 10 item list recall task. The additional variance explained in
the LBLS combined measure o f list and text recall suggests a need for further research on
the advisability and feasibility o f including a measure of text recall in epidemiological
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
research. The drawback of text recall is its considerably more taxing scoring system.
This issue might be overcome by a computer-assisted scoring systems derived from the
traditional Meyer’s (1978) system.
If. on the other hand, investigators are not only interested in gauging memory
ability, but also in issues surrounding the self-perception o f memory ability, self-reported
memory ability could provide a useful source o f information. The subjective beliefs that
people hold about the situation in which they find themselves are powerful forces
regardless of the accuracy of the perception (Cutler and Grams, 1988). Thomas and
Thomas (1928) express the power of the definition o f the situation in their assertion that
"if men define situations as real, they are real in their consequences” (p.572). With regard
to memory ability. Kinsboume (1980) suggests that a low memory self-concept can result
in self-limitation that impairs one’s performance. The findings presented here indicate
that most older adults do not consider their memory abilities to be especially poor which
may portend less self-limitation. An understanding o f self-reports of memory ability
could provide valuable to clinicians who are frequently faced with older patients who
complain o f memory problems. Because of the prevailing negative expectations of
memory loss in old age. self-reports o f memory problems by older persons can be
mistakenly assumed to indicate age decrements in instances when they actually reflect a
more transient conditions such as depression (Zarit, 1981). or medication use (Zelinski &
Gilewski. 1988): or a more fundamental conditions such as neuroticism (Zelinski et al..
1990). or the tendency to complain (Hultsch et al., 1988).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
124
In conclusion, the growing interest in the cognitive functioning o f America's
seniors has moved research on memory ability from the exclusive domain o f cognitive
research into the multidisciplinary arena o f epidemiological research. If carefully
approached, this arena provides increased opportunity for multidisciplinary scholarship, if
not. it provides increased opportunity for inaccurate generalizations. The results o f this
research expose a potential area for such inaccuracy by revealing that unlike seIf-reported
health, self-reported memory ability is not an accurate predictor o f its objective
counterpart. Future research is needed to determine the optimal survey measure of
objective memory performance and the proper interpretation o f self-reports o f memorv.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
125
REFERENCES
Bassett. S.S. & Folstein, M.F. (1993). Self-reported memory, memory performance, &
psychiatric diagnosis: A community study. Journal o f Geriatric Psychiatry &
Neurology, 6, 105-111.
Bazargan. M. & Barbe. A.R. (1994). The effects o f depression, health status & stressful
life-events on self-reported memory problems among aged blacks. International
Journal o f Aging & Human Development, 38(4), 351-362.
Berry. J.M. (1986). Memory complaint & performance in older women: A self-efficacy &
causal attribution model. Unpublished doctoral dissertation. Washington
University.
Bolla. K..L. Lindgren. K..N., Bonaccorsy, C. Bleecker, M.L. (1991). Self-reported memory
in older adults: Fact or fiction? Arch. Neurology 48, 61-64.
Bolla-Wilson. K. & Bleecher. M.L. (1986). Self-reported memory: Dementia or
depression? The Gerontologist, 26, 142.
Burt. D.B.. Zembar. M.J., & Neiderhe, G. (1995). Depression and memory impairment: A
meta-analysis of the association, pattern, and specificity. Psychological Bulletin.
117(2). 285-305.
Cavanaugh. J.C. (1986). Age differences in adults’ self-reports o f memory ability: It
depends on how & what you ask. International Journal o f Aging Human
Development, 24, 271-277.
Cavanaugh. J. & Murphy. N. (1986). Personality & metamemory correlates of memory
performance in younger & older adults. Educational Gerontology, 12. 385-394.
Cavanaugh. J.C. (1989). The importance of awareness in memory aging. In L.W. Poon.
D.C. Rubin. & B.A. Wilson (Eds.), Everyday Cognition in Adulthood & Later
Life. Cambridge: Cambridge University Press.
Cavanaugh. J.C.. & Perlmutter. M. (1982). Metamemory: A critical examination. Child
Development. 53. 11-28.
Christensen. H. (1991). Tire validity o f self-reported memory by elderly persons.
International Journal o f Geriatric Psychiatry, 6, 307-312.
Cole. T.R. (1995). What have we "made” of aging? Journal o f Gerontology, 50(6). S341-
S343.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
126
Costa. P.T.. Jr.. & McCrae, R.R. (1989). Personality, stress, & coping: Some lessons
from a decade of research. In K.S. Markides & C.L. Cooper (Eds.), Aging, stress,
social support & health, 267-283. New York: Wiley.
Costa. P.T.Jr.. & Fozard, J. (1978). Remember the person: Relations o f individual
difference variables to memory. Experimental Aging Research, 291-304.
Cutler S.J.. Grams A.E. (1988). Correlates o f self-reported everyday memory problems.
Journal o f Gerontology, 43, S82-S90.
Deptula. D.. Singh. R. & Pomara, N. (1993). Aging, emotional state. & memory.
American Journal o f Psychiatry, 150, 429-434.
Dixon. R.A. & Hultsch, D.F. (1983a). Metamemory & memory for text relationships in
adulthood: A cross-validation study. Journal o f Gerontology, 38, 689-694.
Dixon. R.A. & Hultsch. D.F. (1983b). Structure & development o f metamemory in
adulthood. Journal o f Gerontology, 38 682-688.
Dixon. R.A.. & Hultsch, D.F. (1984). Metamemory in Adulthood (MIA) Instrument.
Psychological Documents, 14, 3.
Dixon. R.A. (1989). Questionnaire research on metamemory & aging: Issues o f structure
& function. In L.W. Poon, D.C. Rubin, & B.A. Wilson (Eds.), Everyday
Cognition in Adulthood & Later Life. Cambridge: Cambridge University Press.
Dunlosky. J. & Nelson, T.O. (1994). Does the sensitivity of judgements o f learning to the
effects of various study activities depend on when the judgements o f learning
occur? Journal o f Memory and Language, 33(4), 545-546.
Erber. J.T.. Szuchman, L.T.. & Rothberg, S.T. (1990). Everyday memory failure: Age
differences in appraisal & attribution. Psychology & Aging, 5. 236-241.
Flavell. J.H. (1971). First discussant's comments: What is memory development the
development o f? Human Development, 14, 272-278.
Flicker. C.. Ferris. S.H. Reisberg. B. (1993). Longitudinal study o f cognitive function in
elderly persons with subjective self-reported memory. Journal o f the American
Geriatric Society, 41: 1029-1032.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
127
Gilewski. M.J. & Zelinski, E.M. (1986). Questionnaire assessment o f self-reported
memory. In L.W. Poon, T.Cook, K.L. Davis, C. Eisdorfer, B.J. Gurl&, B C .W .
Kasniak. & L.W. thompson (Eds.), Handbook fo r Clinical Memory Assessment o f
Older Adults. Washington, D.C.: American Psychological Association.
Gilewski. M.J.. & Zelinski. E.M. (1988). The Memory Functioning Questionnaire.
Psychopharmacology Bulletin, 24: 655-670.
Hanninen. T.. Reinikainen, K.J., Helkaia E.L., Koivisto, K. Mykkanen. L.. Laakso. M.
Pyorala. K., & Reikkinen, P.J. (1994). Subjective self-reported memory &
personality traits in normal elderly subjects. Journal o f the American Geriatric
Society, 42. 1-4.
Heithoff. K .. (1995). Does the ECA underestimate the prevalence o f late-life depression?
Journal o f the American Geriatrics Society, 43: 2-6.
Hertzog. C. (1992). Improving memory: The possible roles o f metamemory. In D.
Herman. H. Weingartner, A. Searlman, & C. McEvoy (Eds.) Memory
Improvement. New York: Springer-Verlag.
Herzog. R. A. (1995). Approaches for measuring cognitive functioning in large-scale
surveys: Review o f the literature & analysis o f data from the AHEAD survey.
Paper prepared for the AHEAD Workshop, Ann Arbor, MI.
Herzog. R. A. (1994). The measures of cognitive functioning in the AHEAD survey.
Paper prepared for the AHEAD Early Result Workshop. Ann Arbor. MI.
September 1994.
Herzog. R.A. & Rodger, W.L. (1989). Age differences in memory performance and
memory ratings as measured in a sample survey. Psychology and Aging, 4 (2).
173-182.
Hultsch. D.. Hertzog. C.. Dixon, R. & Davidson, H. (1988). Memory self-knowledge &
self-efficacy in the aged. In M.L. Howe & C.J. Brainerd (Eds.), Cognitive
development in adulthood: Progress in cognitve development research. New
York: Spring-Verlag..
Hultsch. D.F.. Hertzog. C.. & Dixon, R.A. (1987). Age differences in metamemory:
Resolving the inconsistencies. Paper in the Techniqcal Report Series o f the
Collaborative Research Group on Cognitive Aging.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
128
Hurd. M.. Rodgers. Soldo, B., Wallace, R. (1994). Asset & health dynamics among the
oldest old: an overview of the survey. Paper prepared for the AHEAD Early
Result Workshop, Ann Arbor, MI, September 1994.
Jutagir. R. (1994). Psychological aspects o f aging: When does memory loss signal
dementia? Geriatrics, 49: 45-53.
Kahn. R.L.. Zarit. S.H. Hilbert, N.M., & Niederehe, G. (1975). Memory complaint &
impairment in the aged: The effect o f depression & altered brain function.
Archives o f General Psychiatry, 32:1569-1573.
Kaszniak. A.W. & Christenson, G. (1994). Differential Diagnosis of Dementia and
Depression. In Storandtt, M. & VandenBos, G.R. (Eds.) Neuropsychological
Assessment o f Dementia & Depression in Older Adults: a Clinician's Guide.
Washington DC: American Psychological Association.
Kaszniak. A.W.. Poon L.W., Reige, W. (1986). Assessing memory deficits: An
information processing approach. In L.W. Poon Ed. Clinical Memory Assessment
O f Older Adults. Washington DC: American Psychological Association.
Kessler. R.C.. Foster, C., Webster, P.S., & House, J.S. (1992). The relationship between
age & depressive symptoms in two national surveys. Psychology & Aging, 7.
H9-126.
Klatzky. R.L. (1984). Memory & awareness. New York: Freeman.
Larrabee. G.J.. & Levin. H.S. (1986). Memory self-ratings & objective test performance
in a normal elderly sample. Journal o f Clinical Experimental Neuropsychology, < S .
275-284.
Lawton. P.. Kleban. K.. & Dean, J. (1993). Affect & age: Cross-sectional comparisons of
structure & prevalence. Psychology & Aging 82, 165-175.
Martin. M. (1986). Aging & pattern of change in everyday memory & cognition. Human
Learning, 5. 63-74.
McCrae. R. R.. & Costa. P.T., Jr. (1987). Validation o f the five factor model of
personality. Journal o f Personality & Social Psychology, 52. 81-90.
O'Hara. M.W. Hinrichs. J.V., Kohout. F.J. Wallace, R.B. & Lemeke. J.H. (1986). Self-
reported memory & memory performace in the depressed elderly. Psychology A -
/l^/ng. /. 208-214.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
129
Perlmutter. M. (1978). What is memory aging the aging of? Developmental Psychology.
14. 330-345.
Poitrenaud. J., Malbezin, M. & Guez, D. (1989). Self-rating & psychometric assessment
of age-related changes in memory among young-elderly managers. Developmental
Neuropsychology, 5, 285-294.
Poon. L.W. (1985). Differences in human memory with aging: Nature, causes. & clinical
implicatons. In J. Birren & K..W. Schaie (Eds.), H&book o f the Psychology o f
Aging 2nd Edition. New York: Van Nostr& Reinhold.
Reige. W.H. (1982). Self-report & tests o f memory aging. Clinical Gerontologist. 12
23-36.
Scogin. F.. & Rohling, M. (1985). Cognitive processes, self-reports o f memory
functioning. & memtal health status in older adults. Journal o f Health & Aging, I.
507-520.
Taylor. J.L.. Miller, &Tinklenberg, J.R. (1992). Correlates o f memory decline: a 4-year
longitudinal study o f older adults with self-reported memory. Psychology &
Aging, 7 185-193.
Theide. K.W. Sc Dunlosky, J. (1994). Delaying students metacognitive monitoring
improves their accuracy in predicting their recognition performance. Journal o f
Educational Psychology, 86(2), 290-302.
Usala. P. & Hertzog. C. (1989). Measurement of affective states in adults: Evaluation of
an adjective rating scale instrument. Research on Aging, 114. 403-426.
Weissman. M. Prusoff, B.,& Newberry, P. (1975). Comparison o f the CES-D with
standardized depression scores at three points in time. Report from from the Yale
University School o f Medicine, SCAN Microfiche No. CCF 000946.
West. R.L. Boatwright. L.K. & Schleser, R. (1984). The link between memory
performance, self-assessment. & affective status. Experimental Aging Research,
10. 197-200.
Williams. J.M.. Little. M.M.. Scates, S., & Blockman, N. (1987). Memeorv complaints
and abilities among depressed older adults. Journal o f Consulting Clinical
Psychology, 55. 595-598.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
130
Yesavage. J.A. (1986). The use o f self-rating depression scales in the elderly. In L.W.
Poon (Ed.) Clinical Memory Assessement o f Older Adults. Washington D.C.
American Psychological Association.
Yesavage. J.A.. Brink. T.L., Rose, T.L., & Adey, M. (1983). The geriatric depression
rating scale: Comparsion with other self-report & psychiatric rating scales. In T.
Crook. S.H. Ferris, & R. Bartus (Eds.), Assessment in geriatric
psychopharmocolgy. pp 153-165. New Canaan, CT: Mark Powlev Associates.
Zarit. S. (1979). Helping an aging patient to cope with memory problems. Geriatrics. 34:
82-84.
Zarit. S.H.. Cole. K.D.. Guider, R.L. (1981). Memory training strategies & subjective
complaints of memory in the aged. The Gerontologist, 2T. 158-164.
Zelinski. E.M.. Gilewski, M.J., & Schaie, K.W. (1993). Individual differences in cross-
sectional & 3-year longitudinal memory performance across the adult course.
Psychology & Aging, 8: 176-186.
Zelinski. E.M.. & Gilewski, M.J. (1988). Assessment o f memory complaints by rating
scales & questionnaires. Psychopharmacy Bulletin, 24: 523-529.
Zelinski. E.M.. Gilewski. M.J.. & Thompson L.W. (1980). Do laboratory tests relate to
self-assessment of memory ability in the young and old? In L.W. Poon. J.L.
Fozard. L.S. Cermak, D. Arenberg, & L.W. Thompson (Eds.), New directions in
memory & aging: Proceedings o f the George A. Tallanard Memorial Conference,
519-544. Hillsdale, NJ: Erlbaum.
Zonderman. A.. Costa. P., & Kawas, C. (1989). Personality predicts complaints o f benign
memory loss. Neurology, 39 Supplement 1, 194.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
On golden cyberpond: Training America's elderly to use electronic mail.
PDF
The effects of psychological type and interpersonal psychological distance on selling-effectiveness in a service industry
PDF
Grandparents and their adult grandchildren: An analysis of the effects of role transition on intergenerational solidarity over time
PDF
Elderly victims of telemarketing fraud: Demographic, social, and psychological factors in victimization and willingness to report.
PDF
Leadership in restructuring schools: The development of an expanded framework
PDF
Determinants of changes in functioning in a sample of elderly Swedes
PDF
Risk-related differences in neurocognitive responses to an ethanol challenge
PDF
Effects Of Conflict On Adult Children'S Relationship To Parents: A Multi-Dimensional Approach To Parent-Child Conflict
PDF
Effects of working memory and semantic impairments on speech in Alzheimer's disease.
PDF
Mechanisms linking SES to hypertension: Findings from three longitudinal studies among the pre-retirement-age and older populations.
PDF
The private language of families: Bringing us news of ourselves
PDF
Processes of loss and recovery of physical abilities in a longitudinal study of Americans 70 years of age and older
PDF
The relationship between identity style and self-regulated learning
PDF
Motivation to learn and to succeed: A path analysis of the CANE model of cognitive motivation
PDF
Evaluating heaviness: Relative weight in the spoken production of heavy-NP shift
PDF
The social cognitive functioning of gang members
PDF
Investigating Stages Of Information Processing and Attentional Capacity: Contributions Of Startle Eyeblink Modification and Measures Of Signal Detection.
PDF
Definiteness types in Spanish: A study of natural discourse
PDF
Measuring young children's coping responses to interpersonal conflict
PDF
Influence of physical activity, social support, and widowhood on disease incidence, mortality, and functional decline in older adults.
Asset Metadata
Creator
Burnight, Kerry Parker
(author)
Core Title
Using self-reported memory ability to measure memory ability in older adults: A meaningful measure or an inapropriate shortcut?
Degree
Doctor of Philosophy
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
Gerontology,OAI-PMH Harvest,psychology, cognitive,Sociology, general
Language
English
Contributor
Digitized by ProQuest
(provenance)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c17-540837
Unique identifier
UC11349894
Identifier
9720197.pdf (filename),usctheses-c17-540837 (legacy record id)
Legacy Identifier
9720197.pdf
Dmrecord
540837
Document Type
Dissertation
Rights
Burnight, Kerry Parker
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA
Tags
psychology, cognitive